The result has been a democratization of development, much as happened with desktop publishing. The Macintosh and PageMaker made it possible for almost anyone to produce a newsletter or brochure, but they did not turn every user into a professional designer, illustrator, or prepress specialist. AI can now produce working software for people who could never have written it themselves, but generating code is only part of building a reliable, trustworthy system.

Today, if you have a repetitive task that’s annoying to do but doesn’t warrant hiring a developer, you can use ChatGPT, Claude, or Gemini to help you create an AppleScript, shell script, Mac or iPhone app, or even a full-fledged Web app. However, if you’re going to dip your toe into AI-assisted development, you need to be careful. Professional developers know to watch out for numerous pitfalls, but if you’re using an AI, it’s up to you to make sure the AI takes them into account. Also, be aware that you’re still ultimately responsible for what you create with AI.

What Could Go Wrong?

In AI-powered development, there are two broad failure categories: problems with the development agent itself and problems with the finished product.

The AI world has moved past basic chatbots to agents that can look up information, change files, write and execute code, drive apps, and more. That makes them much more capable, but it also creates a situation where they could change or delete data, expose confidential data, or install software that itself creates a vulnerability.

On the other side of the equation is the finished product. If it’s local code, it could run but produce incorrect results, fill up a drive with data, overwrite important data, or cause crashes. A networked app could also generate excessive traffic, leak or expose confidential data, or serve as an entry point into your network.

Take AI Development in Steps

Professional developers aren’t born that way. They start with simple tasks and tackle more involved projects as they gain experience. With AI handling the programming for you, it can be tempting to take on a major app, but it’s safer to start with simple data analysis and local automation tools before moving on to full-fledged apps and networked systems. The process won’t turn you into a professional developer, but you’ll have a greater appreciation for all the things they have to consider.

Read-only Experiments

The best way to get started with AI development is with data analysis that would otherwise require you to build spreadsheet formulas, parse large CSV files, compare exported files, and the like. This may not seem like development, but when you drag the files into a chatbot conversation and ask it to work on them, you’ll notice that it does so by writing and executing one or more scripts, often in Python.

With this sort of AI work, you’re limited only by your data and your imagination. You could analyze an event registration or inventory spreadsheet to find trends, get a human-readable summary of errors in a massive log file, or compare data between two differently formatted CSV files to see which people appear in both.

The reason to start here is that you can easily detect errors, and nothing you do can change the data. Incorrect results may still look convincing, however, so compare them against known examples, verify totals, and spot-check the underlying records.

Local Automation Scripts

The next type of AI development to try is local automation via AppleScript or shell scripts. (If you’re uncertain how to execute a script, remember that you can always ask the AI for detailed instructions.) You could build customized scripts for batch-renaming files, creating future calendar events, manipulating images, and even reformatting HTML files.

However, because scripts can modify and delete data, you need to be more careful here. When possible, have the script create new output rather than modifying the original. Otherwise, work on a copy and ask the AI to:

• Warn you about risks or ambiguities in your request
• Preview exactly what it plans to change
• Act only after you confirm the preview
• Log what it changed
• Record enough information to reverse its actions

Standalone Mac or iOS App

Apple’s App Store contains a vast number of apps, but it’s time-consuming to search for apps that promise to do what you want and test them to see if they actually do. All too often, they don’t.

With AI and Apple’s Xcode development environment, you can now create native apps for all your Apple devices. Needless to say, even setting up Xcode is complicated, but once again, an AI agent can do much of it for you and walk you through the rest of the steps. The hard part is often figuring out where some interface control is located, so don’t be shy about pasting a screenshot into the chat and saying, “I don’t see that control—where is it?”

The sky is the limit when it comes to developing your own apps for personal or internal use. You could create an injury rehab tracker, a publishing production checker, a custom document converter, a specialized camera app that adds metadata to photos, and more. Any workflow that’s awkward or doesn’t meet your needs is a candidate. Mac apps can be shared by copying; iPhone and iPad apps must go through Apple’s TestFlight. Again, ask for help.

When it comes to safety, as long as the apps are for personal or internal use, you mostly need to focus on app reliability and data integrity. Make sure to emphasize that the AI should build and run automated tests on every change, and when you’re testing, comment on anything that seems wrong—don’t settle. Specify that data integrity is paramount, require automatic backups, and allow for manual exports.

The bar gets much higher if you plan to distribute the app more widely. You’ll need to consider your obligations regarding support, compatibility, privacy, distribution, and long-term maintenance. An AI can help with some of those issues—be sure to ask it if there are any ways that user privacy could be abused, for instance—but the buck stops with you.

If you’re contemplating selling your app, you’ll also need to confirm that the licenses for any external code, images, fonts, and other components permit commercial distribution. Be extremely cautious if you plan to build a business around an app!

Networked Systems

With great power comes great responsibility. You must be much more careful with networked systems, especially if they’ll be accessible over the Internet. But it can be compelling to build such systems—you could make an equipment checkout system, an event registration system that goes beyond Google Forms, a shared inventory system, a document submission service, or a dashboard that aggregates live data from multiple services.

The problem is that the list of considerations goes well  beyond app reliability and data integrity, including these questions:

• What data and operations must be protected? Some data may be confidential, and personally identifiable information is especially important to protect. Authentication credentials, API keys, and other secrets should never be embedded in source code, prompts, logs, or shared configuration files. Instead, store them in a system designed to manage secrets and limit who and what can access them.
• Who may perform each action, and how is that enforced? Define user roles and limit each role’s permissions to only what is necessary. You don’t want a novice support representative accidentally deleting the database.
• How do you contain hostile inputs and abusive use? Attackers will probe what happens when they submit malicious URLs, API requests, filenames, uploaded files, Web hooks, and other unexpected input. To reduce your exposure, validate all input, limit its size and complexity, and restrict how frequently users can trigger operations.
• How do you separate development, testing, and production? Once you go live, it’s essential to separate these different environments so mistakes made during development don’t take down the live system and data generated during testing doesn’t contaminate the production database. Make sure to separate data and credentials as well.
• How will the system handle simultaneous requests, retries, and partial failures? Network requests can arrive twice, overlap, or fail midway through multi-step operations. Work with the AI to prevent these sorts of issues, which can result in duplicate payments, registrations, messages, or inconsistent records.
• How will you detect and contain attacks, outages, and unexpected behavior? Decide what to log, what conditions should trigger alerts, how access can be revoked, and if it’s possible to isolate a compromised component without shutting down the entire system.

Finally, assign an owner before putting any significant networked system into use. That person must be responsible for investigating failures, restoring data, updating dependencies, renewing credentials and certificates, responding to vulnerabilities, and eventually retiring the system. An internal application that nobody maintains may become less reliable and less secure with every operating-system update, expired credential, and newly discovered vulnerability.

(Featured image by iStock.com/Boonyakiat Chaloemchavalid)

Social Media: Thinking about using AI to write code? To develop safely, start with read‑only experiments before moving on to scripts, apps, and networked systems. Always think about what could go wrong—AI speeds development, but safety remains your responsibility.

Although higher-memory configurations are particularly affected, there’s no guaranteed way to predict which Macs will be affected or the estimated length of delivery delays—the situation changes daily. As of August 4, the Apple Store is reporting the following representative delivery estimates to ZIP code 14850:

• MacBook Neo (8 GB / 256 GB): 10–14 days
• 13-inch MacBook Air (16 GB / 512 GB): 15–22 days
• 14-inch MacBook Pro (16 GB / 1 TB): 1 day
• 14-inch MacBook Pro (24 GB / 1 TB): 8–15 days
• 14-inch MacBook Pro M5 Max (128 GB / 2 TB): 50–57 days
• 16-inch MacBook Pro (24 GB / 1 TB): 1 day
• 24-inch iMac (16 GB / 256 GB): 2 days
• Mac mini M4 (16 GB / 256 GB): 22–29 days
• Mac Studio M4 Max (64 GB / 512 GB): 56–70 days

If you’re in the market for a new Mac, what should you do? It depends on whether you’re buying for yourself or for an organization.

Individuals

If you just want to upgrade from an older model, waiting might not be a hardship—what’s another few weeks? But waiting isn’t easy if you’re replacing a Mac that has been lost, stolen, or broken. You have a few options:

• Shop around: Some retailers may have precisely what you want in stock. Inventory may even vary by Apple Store, if there are multiple stores nearby. It’s often best to call before visiting since online inventory systems may not be up to date.
• Adjust your desired configuration: Standard base configurations often have the best availability. Delivery estimates change with configuration, so if you can get by with less memory or storage, you might be able to get a Mac sooner. Even a different color might be more readily available.
• Choose a different model: As with configuration, some models can be delivered sooner than others, so it may be worth getting an iMac instead of a Mac mini if that means you can have it right away.
• Buy used: Apple sells refurbished Macs (with full warranties) that may have been store models, returned within the 14-day window, or used for demos. Other retailers may also sell older reconditioned Macs, and you can always look on eBay or Craigslist. Thanks to the performance of Apple silicon, a previous-generation Mac isn’t much of a compromise, but be certain any used Mac has Activation Lock removed, isn’t assigned to an organization’s device management system, and is completely erased.

Organizations

The situation may be more problematic for organizations that can’t afford for an employee to be without a Mac or need to equip a new location or a cohort of new hires. Along with the strategies above, organizations should consider:

• Maintaining a loaner pool: Keep a few Macs on hand as backups, even older models that would otherwise have been traded in. This is a good strategy in general, but make sure to refresh the loaner pool periodically so they remain supported and useful.
• Extending replacement cycles: Although many organizations prefer to replace Macs on a 3–5-year cycle, it may be worth extending replacement dates until new Macs can be acquired.
• Handing down equipment: Depending on what’s available, it might make sense to buy an existing employee a new Mac and hand their old Mac down to a new hire.
• Ordering earlier: Build Mac ordering into the hiring process earlier so machines are available when they’re needed.
• Building a deployment reserve: Organizations with predictable hiring may want to keep a few standard Macs ready for immediate deployment. Rotate those machines into normal service rather than accumulating sealed inventory.

Just-in-time manufacturing and overnight shipping have spoiled us, but with some planning and flexibility, you should be able to find the Macs you need to stay productive. Contact us if you need help forecasting Mac purchases, identifying acceptable alternatives, or adjusting deployment workflows around constrained availability.

(Featured image generated by Adam Engst with ChatGPT)

Social Media: Global memory shortages caused by AI data center demand are hitting Mac buyers with higher prices and longer delivery times. Here are strategies to reduce the pain for both individuals and organizations.

This sounds like science fiction—it’s not far off the fictional Kobayashi Maru test in the Star Trek universe, where Captain Kirk beats an unwinnable simulation by secretly reprogramming it—but it’s really a wake-up call about how sophisticated cyber threats have become and why it’s more important than ever before to keep apps, operating systems, and networked devices updated.

OpenAI Models Hack Hugging Face

OpenAI was running internal tests to measure how well its latest models—GPT-5.6 Sol and an unreleased system—performed on cybersecurity tasks. The models were placed in an isolated sandbox environment with their safety guardrails deliberately turned off so researchers could assess their raw capabilities.

The test presented security challenges that the models were supposed to solve. Instead, they decided it would be easier to find the answers—an approach called “reward hacking.” The models spent substantial computing resources searching for a way out of the sandbox, eventually discovering a previously unknown vulnerability in a software proxy that was only supposed to let them download code packages. From there, they worked their way through OpenAI’s internal network until they reached a system with Internet access. (Yes, this level of sandboxing was a mistake on OpenAI’s part.)

Once online, the models reasoned that Hugging Face—a popular platform for AI research—might host the test answers. So they hacked it. They chained together multiple exploits, including stolen credentials and more zero-day vulnerabilities, to breach Hugging Face’s servers. Hugging Face’s security team logged approximately 17,000 hostile events before containing the intrusion.

To defend itself, Hugging Face relied on another AI—ironically, a Chinese open source model, since when Hugging Face tried to work with Claude, its safety guardrails blocked the security-related requests. The open source model was able to analyze the attack, processing the massive event logs quickly so the company’s security team could understand what was happening, stop the attack, scrub systems of the attack code, and deploy additional safeguards.

AI Capabilities Cut Both Ways

Just as human hackers fall into “black hat” and “white hat” camps based on whether they’re attacking or defending, the capabilities that enable AI models to hack their way to a test answer are also what make them increasingly useful for finding and fixing security vulnerabilities.

Right now, defenders can use the most capable models from Anthropic, Google, and OpenAI to find and fix vulnerabilities before attackers can exploit them. But attackers have access to AI models with similar capabilities. Security researchers track what’s called the Zero Day Clock: the time between a vulnerability being discovered and being exploited in the wild. That window has collapsed from months to less than a day, and some predict it will shrink to minutes by 2027.

You can see the defenders at work in the macOS security release notes. In the security release notes for macOS 26.5 and 26.6, the identification of several vulnerabilities was credited to “with Claude, Anthropic” or “Calif.io in collaboration with Claude and Anthropic Research.”

OpenAI’s inadvertent attack wasn’t malicious—its models were simply pursuing the goal of getting a high score on a benchmark without any sense of appropriate boundaries. (In fact, in the short time since the initial OpenAI attack, additional reports of cybersecurity evaluation models hacking into real companies have surfaced from Anthropic, the UK AI Security Institute, Meta, and OpenAI again.) But the technical capabilities these attacks have demonstrated are now part of the threat landscape, available to actors with far worse intentions. It’s only a matter of time before a hostile nation-state gives a capable open source model unlimited computing resources and tasks it with gaining access to secure systems. It’s probably already happening in the cyberdark.

Protect Yourself with Updates

The single most important thing you can do is keep your devices and software up to date. All those security patches from Apple, Microsoft, and app developers address vulnerabilities that AI-powered tools can find and exploit. Worse, once these vulnerabilities have been disclosed, hackers may be able to quickly leverage general knowledge of them to fabricate and deploy attacks against those who haven’t yet updated.

Our advice continues to be:

• Enable automatic updates: Don’t allow yourself to forget to install updates. On iPhones and iPads, go to Settings > General > Software Update > Automatic Updates. On Macs, go to System Settings > General > Software Update, click the ⓘ button next to Automatic Updates, and turn on all the switches. (If you’re working in an organization with an update policy, check with IT first.)
• Keep apps updated: Your apps need updates too, including your Web browser, email client, messaging client, and any software that touches the Internet. Attackers often target the weakest link—an outdated app can provide an entry point even if your operating system is current.
• Update device firmware: It’s equally important to update the firmware on routers, switches, printers, and other networked devices that could serve as entry points for attacks or be recruited into a botnet.
• Consider security in hardware upgrades: Although the main reason to upgrade Apple hardware should be functional, keep in mind that a newer device will likely be more secure thanks to improved hardware protections.
• Replace unsupported devices: Older hardware that no longer receives security updates should be upgraded, whether it’s a Mac or iPhone, network hardware, or a device like a printer, security camera, or Internet-connected doorbell.

With sufficient attention from developers, security may eventually become less important for us to watch than it is today. But things will get worse before they get better, so please excuse us as we keep trying to get everyone to install updates regularly.

(Featured image by iStock.com/Thinkhubstudio)

Social Media: AI models hacked a real company while trying to cheat on a test—escaping their sandbox, finding unknown vulnerabilities, and breaching production servers. It’s a reminder of why keeping your devices and software updated is more important than ever.

The chat interface that today’s AI systems still rely on has become more of a control panel than the system itself. What happens after you press Return may involve Web searches, file analysis, code execution, connected accounts, and even digital-world actions—all orchestrated behind the scenes. You need to understand what’s happening behind the chat box to evaluate the accuracy, quality, and utility of the answers that appear there.

Under the Hood of an AI System

The capabilities of a modern AI system include:

• Models generate text, analyze images, and work through problems. This is the traditional “AI” part, but it has improved hugely through better training techniques, longer conversations and documents, and extended reasoning time. Cutoff dates for training models remain an issue—some AIs think macOS 26 is still in beta.
• Retrieval pulls information from the Web, uploaded documents, or connected data sources. When an AI cites a source, it’s usually because retrieval happened. However, it’s still essential to check cited sources carefully because the URLs may be broken, and even when pages exist, they may not support the claims.
• Tools handle tasks the model can’t do on its own, such as analyzing numeric data, running code, creating visualizations, or searching databases. For many types of requests, the AI will write a script or call a tool to handle the work, dramatically increasing the likelihood that the results will be correct and making it easier to refine them.
• Connectors link AI systems to external platforms such as email, calendars, file storage, and Internet-hosted applications. They’re necessary so a system can work on personalized data (“Give me a timeline of the deliverables on the MacDavis project from our email conversations.”) and to connect with business data.
• Actions let AI systems do things in the digital world: send messages, create events, modify files, and interact with other software. Here is where AI stops merely advising and starts affecting real systems, so the safeguards need to be much stronger.

These capabilities don’t always appear together. A simple chatbot exchange may rely only on the model’s training. A research request may add retrieval and tools so the AI can search current sources, summarize what it finds, and run calculations or create charts. A workplace copilot may add connectors to email, calendars, cloud storage, customer records, or internal databases. A full-fledged agent adds actions, enabling the system to operate on your behalf.

It’s important to understand all the possibilities because each layer changes both what the AI can do and how much you should trust it. A model-only answer calls for skepticism (and perhaps a search). A search-based answer needs source checking. A tool-generated answer requires checking the inputs, method, and results. A connector-based answer warrants attention to the source of the data (and whether the permissions are too broad). And an action deserves a preview, an approval process, and ideally a way to undo mistakes.

What This Means for Trust, Privacy, and Control

This evolution from chatbot to assistant generally yields better results, but it also comes with new risks:

• Privacy isn’t just about what you type. It’s about what files you upload, what accounts you connect, and what the AI system can access. Putting confidential data into an AI could be problematic in numerous ways, not least regulatory compliance.
• Accuracy varies by task. Responses drawn purely from training data may be outdated or flat-out wrong, while search-informed answers reflect current sources. Asking an AI to “Confirm with a search” is a good way to get it to reassess what it has written with current information.
• Actions have consequences. When AI can send email, modify files, change settings, or interact with business systems, mistakes can have outsized impacts. This is why we strongly recommend caution about OpenClaw-style agents: giving any AI broad access to email, messaging, calendars, files, and applications creates security and reliability risks. Always preview any action—or AI-generated script you’re running externally—whose results could be difficult or impossible to reverse.
• Confidence isn’t correctness. AI outputs can sound polished and authoritative even when they contain errors, outdated information, or subtle misunderstandings of your intent. Any AI-generated work that informs decisions or will be seen by others deserves human review. Just as you wouldn’t base anything important on the work of a summer intern without checking first, you don’t want AI-generated work to go out if you can’t stand behind it.

For individuals, how you react to these risks mostly comes down to verification and restraint: check important sources, review important outputs, and don’t let AI take irreversible actions without approval. For organizations, the same principles must become policy because employees may already be using AI tools with company data without IT’s knowledge. To get ahead of the issue, organizations should:

• Audit what’s already happening. Talk with employees to find out what they’re already using, inventory online apps with embedded AI features, and check browser extensions, which often fly under the radar.
• Classify workflows by risk. Not all workflows need the same level of attention. Low-stakes tasks like brainstorming and text editing can proceed with commonsense guidelines. High-stakes outputs—client communications, financial analysis, security configurations—need human review before they go live.
• Evaluate tools carefully. Before approving any AI tool for general use, understand what data it accesses, whether it trains on your inputs, what actions it can take, and what logging is available. Most paid tools won’t train on your data, or can at least be configured not to, but the more important your data, the deeper your research should go.
• Create clear policies and train employees. Define which tools are approved, what data can be entered into AI systems, when human review is required, and what’s prohibited. Employees need practical guidance on how to apply the policies to their actual workflows. Don’t assume that an email or two is sufficient—training is essential.

The biggest mistake people make about AI today is underestimating both its risks and rewards because they’re still thinking about ChatGPT from 2023. The chat box may look the same, but it now sits in front of systems that are vastly more powerful, meaning that it’s more important than ever to consider when to trust them, when to verify them, and when to keep them at arm’s length.

(Featured image by iStock.com/tadamichi)

Social Media: Still thinking of AI as just a chatbot? Today’s tools search the Web, run code, access your files, and can even take actions—all triggered from the same chat interface. The results are better but come with new responsibilities.

In April 2026, AI company Anthropic announced Mythos Preview, an AI model with unprecedented capabilities for finding security flaws in software. Unlike previous AI-assisted security tools that often produced false positives and created more work for human evaluators, Mythos proved that it can discover true vulnerabilities that have evaded detection for decades—including a 27-year-old bug in OpenBSD, an operating system famous for its security.

Alongside Mythos, Anthropic launched Project Glasswing, a collaborative effort with approximately 50 partners—including Apple, Microsoft, Google, Amazon, and numerous financial institutions—to identify and fix vulnerabilities in critical software before similar AI-powered capabilities fall into the hands of malicious actors. As Firefox engineers said when reporting on their use of Mythos, “The current moment is a perilous one, but also full of opportunity. Let’s work together to secure the internet.”

Why This Matters

The security implications are significant. Within a month of Project Glasswing’s launch, Anthropic reported that partners collectively found more than 10,000 high- or critical-severity vulnerabilities in their software. Several partners reported that their bug-finding rate increased by more than tenfold.

Firefox developer Mozilla published compelling evidence of Mythos’s capabilities, noting that it identified and fixed 271 vulnerabilities in Firefox 150—over 10 times as many as they found in Firefox 148 with the previous-generation Claude Opus 4.6. Of those 271 bugs, 180 were rated high-severity, meaning they could be exploited through normal user behavior, such as browsing a Web page. (The chart below shows higher numbers because it includes bugs from other sources and other versions.)

Mozilla’s results are remarkable for both their volume and their quality. Firefox engineers reported finding bugs that had remained undiscovered through many years of traditional security testing.

How Apple Fits In

Apple is a founding partner in Project Glasswing, yet another signal that the company takes security seriously. Apple’s vertical integration—controlling everything from chip design to the operating system to the App Store—gives it a structural advantage in secure design. For instance, Apple’s newest M5 Mac chips and A19 iPhone and iPad chips include Memory Integrity Enforcement (MIE), a hardware-level protection designed specifically to stop memory corruption exploits.

However, even these cutting-edge protections aren’t invulnerable. Security researchers at Calif.io demonstrated the first public macOS kernel memory corruption exploit on M5 silicon in May 2026. Working with Mythos Preview, they built a working privilege escalation exploit in just five days—targeting hardware protections that Apple spent five years developing.

An Arms Race Against Time

The uncomfortable reality is that while Anthropic currently controls access to Mythos, equivalent capabilities will inevitably become more widely available. OpenAI and Google probably already have similarly capable models in development, as do other AI model developers, some of whom may be accessible to or even beholden to hostile entities.

The security dynamics differ depending on whether we’re talking about existing software or new code that hasn’t shipped yet. For new code, defenders have a clear advantage—they can scan for vulnerabilities before release and catch bugs that would never have been found manually.

For existing software already running on billions of devices, the picture is darker. Attackers only need to find one exploitable bug to get in; defenders need to find and fix all of them. Worse, attackers don’t have to test their code to avoid breaking features, schedule a release, or get approval from other departments—they can exploit a vulnerability the moment they find it. As the Zero Day Clock site shows, the window between a vulnerability being discovered and being exploited has dropped precipitously—what once took months now happens in days and is expected to happen in minutes in a year or two.

This transition period—while AI rapidly discovers vulnerabilities in existing code that takes time to patch—is where we’re most at risk. The Zero Day Clock site features a call to action offering 10 suggestions for how the industry—and society—should rethink cybersecurity to stave off this threat.

Keep Installing Updates

For most users, the best defense against AI-powered exploits is nothing new—keep your devices updated—but it’s more important than ever. The fixes in Apple’s updates increasingly include patches for AI-discovered vulnerabilities. Practically speaking, you should:

• Enable automatic updates: Don’t allow yourself to forget to install updates. On iPhones and iPads, go to Settings > General > Software Update > Automatic Updates. On Macs, go to System Settings > General > Software Update, click the ⓘ button, and turn on all the switches. (If you’re working in an organization with an update policy, check with IT first.)

• Consider security in hardware upgrades: Although the main reason to upgrade hardware should be functional, keep in mind that a newer device will likely be more secure thanks to improved hardware protections.

• Replace unsupported devices: Hardware that no longer receives security updates is increasingly risky. This applies not just to your Apple devices but to every piece of gear that can be updated, including network hardware, printers, and smart home devices like cameras and doorbells.

In the long run, the emergence of AI-powered security tools favors defenders—developers will be able to catch many more bugs before shipping. For the near future, however, it’s essential that we keep our devices running the latest and most secure software.

(Featured image by iStock.com/Thinkhubstudio)

Social Media: Anthropic’s Mythos Preview AI model found over 10,000 high- or critical-severity vulnerabilities in one month. Similar AI models will soon be available to attackers, so keeping your devices updated has never been more important.

(Featured image by iStock.com/demaerre)

Social Media: Apple Music’s new Playlist Playground in iOS 26.4 uses AI to create custom playlists based on any mood, genre, or activity you describe. Just type what you’re in the mood for and let the app do the rest.

That’s the promise driving the recent surge in popularity of OpenClaw (formerly known as Clawdbot and Moltbot), which is now all the rage in tech circles. Token Security found that at least one person is using it at nearly a quarter of its enterprise customers, mostly running from personal accounts. That’s a shadow IT nightmare—employees connecting work email and Slack to an unsanctioned tool that IT doesn’t know about and can’t monitor. Whether you’re an individual tempted by OpenClaw’s promise or a manager wondering what your users are up to, you need to understand the risks these AI agents pose.

OpenClaw is an AI agent built around “skills”—installable plugins that let it integrate with your messaging apps, email, calendar, and more. You communicate with OpenClaw via Messages, Slack, WhatsApp, and similar apps. Because it’s open source, you’ll need to provide your own API keys for AI services like OpenAI or Anthropic, which means ongoing costs that can add up quickly—people have reported spending $10–$25 per day.

The more serious problem? Security researchers have discovered serious vulnerabilities, including misconfigured instances exposed to the internet that leak credentials, API keys, and private messages, and a supply chain vulnerability where malicious skills uploaded to the ClawdHub library can execute arbitrary commands on users’ systems. Even beyond specific bugs, OpenClaw’s fundamental design encourages users to grant broad access to sensitive accounts.

Why AI Agents Are Risky

Security concerns aren’t unique to OpenClaw—they apply to any AI agent that acts on a user’s behalf. Here’s what’s at stake:

• Credential exposure: For an AI agent to send emails, manage your calendar, or post to Slack, it needs your authentication tokens or login credentials. If the agent software stores these credentials insecurely, or an attacker gains control, they could be exposed.
• Prompt injection: AI agents work by following instructions, but they can’t easily distinguish between prompts and data in the content they use. A class of attacks called “prompt injections” trick AI systems by hiding malicious content in emails, websites, or documents that will be processed. An attacker could embed instructions in an email that would cause your agent to search for and forward email messages containing passwords or financial data, follow links to malware sites, or take other harmful actions. There is currently no foolproof defense against this class of attack.
• Data exfiltration: An AI agent with access to your email and your computer’s filesystem could be manipulated to extract information from elsewhere on your computer—financial data, customer lists, or personal details—and send it to an attacker.
• Unvetted extensions: OpenClaw and similar AI agents let users install “skills” or plugins to extend functionality. Libraries that allow users to share custom skills often have minimal or no security vetting, making it easy for attackers to submit poisoned skills. Installing such a skill could grant malicious code access to everything your agent can touch.
• Exposed control interfaces: Security researchers found OpenClaw control servers exposed on the Internet, potentially leaking API keys, VPN credentials, and conversation histories. This risk is unique to OpenClaw at the moment, but future AI agents may suffer from similar vulnerabilities, particularly as they’re adopted by less technically savvy users.

How to Reduce Your Risk

We’ll come right out and say it: we strongly recommend against installing OpenClaw or other AI agents on your Mac. In a year or so, Apple may have updated Siri to provide many of these capabilities with significantly stronger privacy and security. But for now, just say no.

If you decide to use AI agents despite these risks, here are practical steps to protect yourself:

• Use dedicated accounts: When possible, create separate accounts specifically for agent use rather than linking your primary personal or work accounts.
• Limit permissions: Grant the agent access only to accounts it absolutely needs. If you only want help with your calendar, don’t also connect your email and messaging services.
• Avoid connecting sensitive services: Never connect anything involving money, healthcare, or confidential business information. The liability is too high if something goes wrong.
• Review agent actions: If the platform offers logs or activity feeds, check them regularly. Look for unexpected messages sent, files accessed, or connections made.
• Vet extensions carefully: Don’t install skills or plugins from unknown sources, and even with known libraries, look for evidence of others using and reviewing the skills. Treat skills like any other software you’d install on your computer.
• Keep software updated: Security patches for OpenClaw and similar tools address known vulnerabilities. If you’re running an agent, keep it up to date.
• Run agents in isolated environments: Technical users should consider running agents in sandboxed environments or virtual machines to limit potential damage.

If you run a business, you should assume that some employees have already installed OpenClaw or will soon, and may have connected their work email and Slack accounts without realizing the associated risks. Here’s what you can do:

• Educate before it’s a problem: Proactively explain the risks to employees. People are more receptive before they’ve already invested time setting something up.
• Update acceptable use policies: Make clear that connecting work accounts to unsanctioned AI agents is prohibited, and explain why.
• Offer sanctioned alternatives: If employees want AI assistance, point them toward safer options that don’t require handing over credentials to sensitive accounts.

What About Claude Cowork and OpenAI Codex?

Not all AI agent platforms carry the same level of risk. Anthropic’s Claude Cowork and OpenAI’s Codex take a different architectural approach from OpenClaw. Rather than requesting authentication tokens for your email, messaging, and other personal services, they operate within their own controlled, sandboxed environments. These systems work primarily with files, code, and data you explicitly place into their workspace, which substantially limits the fallout from an attacker gaining some level of control.

This containment approach reduces risk, but does not eliminate it. Prompt injection remains a concern whenever an AI system processes untrusted content, even inside a sandbox. An AI agent analyzing a malicious document could still be manipulated into taking unintended actions within its allowed environment. Similarly, any code generated by these systems—particularly code that touches the network or executes system commands—should be reviewed carefully to make sure it hasn’t been compromised by prompt injection.

The key distinction is scope. Claude Cowork and Codex are designed to operate within a defined workspace, whereas tools like OpenClaw require standing access to your most sensitive accounts. From a security perspective, a compromised sandbox is a recoverable incident; a compromised email or messaging account may not be.

The Bottom Line

AI agents promise a lot and may provide genuine convenience, but at a cost beyond just paying for API tokens. Before you or anyone in your organization connects an AI agent to sensitive accounts, consider: What’s the worst that could happen if this system were compromised by an attacker? If the answer involves passwords being stolen, private email being exposed, or photos being posted to social media without your knowledge, proceed with extreme caution. If you can imagine a way financial accounts could be accessed or business data stolen, don’t proceed at all.

(Featured image by iStock.com/Thinkhubstudio)

Social Media: AI agents like OpenClaw promise to automate tedious tasks, but recent security vulnerabilities highlight the dangers of using them. Learn the risks and how to protect yourself—and your organization—if you choose to use an agent.

As always, reality lies between the extremes. AI is just another technological tool, like spreadsheets, email, and the searchable Web. Like them, casual usage won’t automatically increase an organization’s productivity. At best, many people have begun using AI chatbots as a smarter search engine, and while that’s a fine start, it’s unlikely to make a notable difference. Many others are technology skeptics who are uncomfortable with any new technology, let alone one as fuzzy as AI. Even those who are interested and capable are often overwhelmed by their existing work and don’t have time to learn yet another tool.

So how do you set up an organization to make effective—even transformative—use of AI?

Get Buy-In from Management

Ideally, the desire to adopt AI would come from the top of the organization, with leadership discussing and modeling the kind of usage they want to see. But what’s absolutely essential is lower-level management creating the culture, resources, and time necessary for employees to experiment with AI.

Evangelize from the Bottom, Don’t Mandate from the Top

Although management must be on board, a CEO memo mandating immediate AI adoption won’t have the desired effect. Unlike many other technologies, AI solutions tend to be highly specific rather than one-size-fits-all. Frontline employees know where they’re wasting time with inefficient workflows, and they have first-hand knowledge of what customers want, so they’re more likely to be able to leverage AI tools when they are involved in the development and deployment. Solutions created without their participation likely won’t benefit the business’s bottom line, customers, or employees.

Centralize Testing and Support

A top-down approach does make sense for tool analysis and testing. The explosive growth of the AI market means that there are numerous similar options for any desired workflow. To save time, avoid future chaos, and reduce tool jumping, it can be helpful to have a single IT team evaluate the numerous possible tools, make recommendations, suggest best practices, establish basic data handling and privacy guidelines, and provide support.

Adopt a Documentation Mindset

A key to automating workflows with AI is being able to document the necessary tasks clearly first. Some organizations already have a documentation mindset, where they write everything down, define processes, and record decisions. If that’s not the case for your organization, it’s better to focus on building such documentation before creating automation tools that are unlikely to deliver the desired results. Consider using AI to help with documentation, such as by interviewing people who understand the workflows and using AI to extract an outline from the transcript of the recording.

Think of AI Tools Like a Junior Employee

The hard part of using AI is defining your goals and determining where AI can make a difference. It’s much like training a new hire. What are you trying to achieve by hiring them? What do they need to learn to do their job? What level of excellence do you expect? What common mistakes and pitfalls should they avoid? You can only automate something if you have a clear idea of what success entails and precisely what’s necessary to achieve it.

The Bottom Line

Ultimately, successful AI implementation comes down to defining what you want to achieve, giving people the time they need to explore possibilities, and providing guidance rather than mandates.

(Featured image by iStock.com/FabrikaCr)

Social Media: Casual AI use won’t impact your organization. To see real productivity gains with AI projects, avoid top-down mandates and instead empower frontline teams, document workflows, and centralize support.

(Featured image by iStock.com/tadamichi)

Social Media: Hackers have learned how to poison shared chatbot conversations with malware—and get Google to display them in search results. Never trust step-by-step instructions or Terminal commands from user-generated chatbot pages.

Triggering Visual Intelligence

We offer numerous examples of visual intelligence’s superpowers later in this article, but first, let’s make sure you know how to activate its two modes: camera mode and screenshot mode. Use camera mode to learn more about the world around you; use screenshot mode for help with something on your iPhone’s screen. Here’s how to trigger each mode:

• Camera mode: Press and hold the Camera Control button on all iPhone 16 models (except the iPhone 16e), all iPhone 17 models, and the iPhone Air. Press the Camera Control again or tap the shutter button to lock the image for visual intelligence. On the iPhone 15 Pro, iPhone 15 Pro Max, and iPhone 16e, which support Apple Intelligence but lack the Camera Control, use the Action button, a Lock Screen button, or a Control Center button.
• Screenshot mode: Simultaneously press the side button and volume up button to display an interactive preview of what was on the screen.

Visual intelligence analyzes the content of your image and provides relevant action buttons based on what it detects. While the Ask and Search options are always available, other buttons appear contextually depending on the content.

• Ask: Tapping the Ask button lets you pose a question about the image to ChatGPT. (But we’ve found that Apple won’t pass on health-related questions or queries with certain types of sensitive data.)
• Search: Tapping the search button conducts a Google Image search for similar items. It may also display tabs for search results from other apps, such as Etsy or eBay.
• Recognized objects: When visual intelligence identifies an object, such as a specific plant or animal, it displays a button that brings up more details.
• Text: When it detects blocks of text, visual intelligence provides buttons to summarize the text, read it aloud, or translate it.
• Dates: If it detects a date in text, visual intelligence displays an Add to Calendar button.
• Contact info: When details like email addresses or phone numbers appear in the image, visual intelligence can help you call or message the number, or send email.
• Addresses: When it identifies a physical address in text, visual intelligence displays a button that opens the address in Maps.
• URLs: This works in the virtual world as well—a URL embedded in an image prompts visual intelligence to display a button that opens the website in Safari.
• Businesses and locations: When you capture an image of a business or other location that’s known in Maps, visual intelligence can show hours, menus, reviews, and more.

Real-World Uses for Visual Intelligence

It can be challenging to think of uses for visual intelligence at first, simply because it’s a new way of engaging with the world around you and what you see on your iPhone. We’re used to taking pictures of event flyers we want to attend, doing Google searches for things we see, asking questions of chatbots, and using specialized apps to identify plants and animals—visual intelligence can do all that and more. Here are a few practical ways to use visual intelligence today:

• Create calendar events: Create calendar events from posters, flyers, invitations, or Web pages. When you point the camera at a poster or take a screenshot of an event page, an Add to Calendar button allows you to create an event directly from the on-screen details.
  
• Find business information: Point the camera at a business to retrieve details such as hours, menu/services, phone number, and website. It’s the same information you’ll find in Maps, but it’s easier to pull up using visual intelligence.
  
• Search for products: Shopping for something? Once you find an example of what you like—such as this mid-century modern sofa—take a screenshot, circle the picture with your finger, and browse the search results for similar items.
  
• Summarize and read text aloud: When you’re faced with a large amount of text, especially if the font size is difficult to read, visual intelligence can provide a summary or even read it aloud. The option to have text read aloud can be particularly helpful for those with low vision.
  
• Translate text: iOS offers multiple ways to translate text in unfamiliar languages, including the Translate app, but visual intelligence is often the fastest way to get a quick translation of a sign or placard.
  
• Quick object identification: We’ve all wondered what some plant or animal is—using visual intelligence, you can point your camera at it to find out quickly. Just tap the name that appears to get more details.
  
• Research questions: Sometimes, you know what you’re looking at but have questions about it. Instead of starting a new search, you can use visual intelligence, tap the Ask button, and pose your question to ChatGPT. Tap the response to ask a follow-up question. If you connect Apple Intelligence to your ChatGPT account in Settings > Apple Intelligence & Siri > ChatGPT, your conversations will be saved in ChatGPT, where you can review and continue the discussion.
  
• Manipulate real-world data: Anything that can be photographed or captured in a screenshot can be used as data for other manipulations. For example, you could take a picture of a bookshelf and ask for a list of all the titles, or take a screenshot of a recipe and request the calorie count per serving.
  
  

How does visual intelligence compare to apps like ChatGPT, Claude, Gemini, and others? It outdoes them in two ways but falls short in one. Thanks to its deep integration with iOS and the iPhone’s Camera Control, it’s easier to activate visual intelligence than any other app. It also transfers data more effectively to other apps, such as sending URLs to Safari, phone numbers to Phone or Messages, addresses to Maps, and more. However, chatbot apps—which can also analyze photos and screenshots—are more conversational, offer more detailed information, and are willing to discuss potentially sensitive topics that Apple won’t touch, such as health and politics. We use visual intelligence for straightforward tasks, but for more complex situations, we often turn to a chatbot app instead.

(Featured image by Apple)

Social Media: With its visual intelligence feature, Apple Intelligence brings superpowers to your iPhone’s camera and screen. Just point and hold to identify objects, translate text, create calendar events, and get answers about anything you see.

First off, we’re not offended. If you can work through simple problems on your own with the help of an AI chatbot, that lets us focus on helping you with the bigger picture and issues that can be solved only by someone with awareness of your physical setup or broad knowledge of your workflow.

However, the mere fact that we’re getting these questions shows that people aren’t entirely comfortable with the AI answers, which is a good thing. Because chatbots work by giving you the most statistically likely words from their training models or extracted from search results, they can sometimes return incorrect information that could be harmful or even damaging. And, of course, they’ll do so in a breezy, confident tone that doesn’t suggest any concern.

For instance, we’ve seen chatbots confidently suggest deleting files or resetting permissions from the command line (be very afraid of anything that starts with sudo), disabling System Integrity Protection (almost never necessary), turning off Gatekeeper to install unsigned apps, resetting iCloud Keychain syncing, and more.

Here’s how to think about those responses. First, if you sense that following the chatbot’s instructions might cause problems, ask it to explain potential concerns and how to address them. Also, if you don’t understand what it’s telling you to do, say so and ask it to restate the instructions in simpler terms for someone less experienced. After pushing the chatbot for more details, use your critical thinking skills to ask yourself if its instructions could lead to irreversible changes or data loss.

If you still have any hesitation after going through that process, then it’s time to contact us. It’s helpful to share your chatbot conversation with us so we can assess what it suggested and explain why there was no need to worry or why you were justified in checking before taking action that you might regret.

Although this may seem like a modern problem, we’ve seen many similar situations over the years, where people get fired up about an article they read in an airline seatback magazine or hear something from their brilliant nephew who’s getting a degree in computers from a very good college. Although there’s no intent to deceive from any of these sources (chatbots don’t have intent at all, much less any to deceive), technical advice only makes sense in the context of your goals and resources.

In fact, having conversations about AI suggestions can be helpful because they help you develop better technical judgment. We can help you understand the principles behind different technical solutions, highlight what details matter when evaluating recommendations, and build your confidence in knowing when to trust (or distrust) technical advice from any source. Think of it as collaborative problem-solving that makes you better equipped to handle future technical challenges, whether you tackle them independently or with our professional help.

For the record, chatbots can help you understand basic settings, find features in common apps, and interpret standard error messages. But whenever a suggested solution involves system-level changes or seems risky, that’s when you should contact us.

(Featured image by iStock.com/Valerii Apetroaiei)

Social Media: When should you trust an AI chatbot’s tech advice? Learn when chatbots can help with basic issues and when it’s time to call a professional for more complex problems.

This issue affects all major videoconferencing platforms—Zoom, Microsoft Teams, Google Meet, and others. Many organizations also use standalone AI recording tools that can join meetings as participants, such as Otter.ai, Fireflies.ai, and tl;dv.

No one should feel ashamed of using AI-generated meeting summaries, nor should these tools be categorically avoided. They’re undeniably helpful, allowing people to focus on the discussion instead of taking notes or worrying about forgetting action items. We know people who consider them life-changing.

However, the fact remains: unlike a person tasked with taking notes, these tools record everything, including pre-meeting chatter, small talk, and personal asides that a person would know not to include. Making matters worse, AI notetakers are often configured to distribute transcripts and summaries automatically to all attendees—including those who were invited but didn’t attend. While this helps people catch up on missed meetings, it can cause problems if the absent individuals were themselves the topics of discussion. And we won’t even get into the potential legal and HR implications of certain conversations being made public.

Practical Solutions

Given the utility of AI-generated meeting summaries, what can you do to reduce the chances of potentially embarrassing or problematic conversations being shared inappropriately?

• Warn attendees: Although most videoconferencing tools alert users that recording is happening, everyone is used to these notifications. For a more explicit warning, the meeting host can remind everyone that summaries will be shared with all attendees.
• Pause/resume recording: While not all videoconferencing and AI recording tools offer the option to pause and resume, it can be useful. The meeting host can wait to start recording until everyone has arrived and the pre-meeting chatter has died down, and then stop it once the last agenda item has been discussed. The challenge is that this requires the host to remember to start and stop at the right times, and any valuable conversation before or after these points will be lost.
• Restrict distribution: Another option is to configure the system so meeting summaries are sent only to the host, who can then review and edit them if needed before sharing with the rest of the attendees. The drawbacks here are the extra work for the host and the delay in participants receiving the notes, which can hold them back from starting on action items.
• Watch what you say: Just as with social media posts, it’s important to think before you say something you might regret. If you assume that everything you say could be shared with your entire organization—including HR and your boss—you’ll be much less likely to get into trouble. Of course, this requires everyone to be sufficiently self-aware to avoid problematic topics.
• Use private channels: If you anticipate needing to discuss sensitive information with a remote colleague—the kind of thing you’d shut your office door to keep passersby from overhearing—use a private channel like a personal meeting room, direct message, or phone call. And if someone starts to say something problematic in a group meeting, gently suggest moving it to a private channel.

Although having AI-generated summaries of conversations you thought were private circulated to others may feel like a modern problem, variants have been around for a long time: the romantic message misaddressed to the company-wide email list, the list of layoffs left in the copy machine, or even a conversation that continues across stalls in the bathroom without realizing someone else has come in. Ultimately, all we can do is be mindful of what we say and who might hear it.

(Featured image by iStock.com/ArnoMassee)

Social Media: In virtual meetings, AI recording tools often capture and share everything—even those casual chats that occur after most attendees have left. Learn how to avoid having sensitive conversations broadcast to your whole team.

• ChatGPT: Go to Settings > Data Controls and turn off “Improve the model for everyone.”
• Claude: Navigate to Settings > Privacy and disable “Help improve Claude.”
• Gemini: Visit the Your Gemini Apps Activity page and turn off Gemini Apps Activity.
• Meta AI: Avoid it entirely, as it doesn’t allow you to opt out of training.

(Featured image by iStock.com/wildpixel)

Social Media: Don’t let sensitive business data become part of AI training sets. Here’s how to turn off training options in popular AI chatbots to protect your company’s information.

For those who don’t want to spend time getting to know all the different AI tools, here’s our advice on the best options right now. This isn’t an area to go for an underdog—if you want to get serious work done, stick with one of the major systems. That means choosing from these five, which make up about 90% of the market:

• ChatGPT from OpenAI
• Gemini from Google
• Claude from Anthropic
• Copilot from Microsoft
• Perplexity from Perplexity AI

That’s roughly the order in which we recommend you choose, although there are particular areas where one outperforms the others. Realistically, though, you’ll be fine with any of them, so the decision may come down to which feels the best during your interactions on the Web and through the system’s mobile app.

Although you can test all five chatbots for free, you’ll get much better results if you pay for an account to unlock unlimited access to all models and features. The free accounts use older, less advanced models and will cut you off if you try to use them too much in a day. All these chatbots cost $20 per month, other than the full version of Microsoft 365 Copilot, which is $30 per month and requires an annual commitment.

Let’s explore different usage scenarios that could make one chatbot a better fit for your needs.

General Purpose Usage: ChatGPT and Gemini

If you’re unsure about what you want to do with a chatbot but want to explore all options, the best overall choice is ChatGPT, with Gemini as a close runner-up. Both excel at generating text and images, can include content from real-time Web searches in their responses, offer a voice mode that lets you speak and hear responses, provide a choice between fast and advanced models, and can conduct in-depth research.

Writing and Editing: ChatGPT and Claude

ChatGPT is regarded by many as the top choice for writing, thanks to its fluency across different styles. It provides excellent rewriting and summarizing capabilities, and its GPT-4.1 model is currently the favored option in the AI-enhanced Lex word processor, which merits honorable mention for those who do a lot of writing. Claude also performs well with writing, especially when you’re interested in a deep understanding of structure, tone, and intent. If your writing mostly happens in Microsoft Word and Outlook, the Copilot add-on offers useful integration. Likewise, if you live in Google Docs, you might prefer Gemini for its integration.

Research and Current Information: Perplexity and ChatGPT

Many people think of—and interact with—chatbots as souped-up search engines. That’s not entirely helpful because chatbots can do vastly more than search the Web, but incorporating current information from the Web is essential for many tasks. Although most chatbots can now access the Web, you’ll get the best results from Perplexity and ChatGPT, with Gemini being a close second. Claude had the weakest search in our experience.

Image and Video Generation: ChatGPT

ChatGPT shines here as the front end for OpenAI’s DALL-E 3. What makes it stand out from most other chatbots is its prompt understanding—how well it translates your request into a desired image—and its ability to do “inpainting.” Before DALL-E 3, most artbots refined images by changing the prompt and starting over each time, which often led to frustrating results when all you wanted was a specific change. With inpainting, you can ask ChatGPT to modify an image—such as changing a person’s hair color—without affecting the rest of the picture. Although generating videos with a chatbot is mostly a gimmick right now, using Sora through ChatGPT will also provide the best results.

Voice Mode: ChatGPT and Gemini

One of the most impressive tricks of chatbots is their ability to engage in natural conversations with you about nearly any topic. ChatGPT’s Voice Mode is generally regarded as the best, and its speech can be highly realistic, down to pauses, occasional “um” insertions, and varied inflection. Gemini is also good, and both are noticeably better than Claude at this time. When you use voice mode in the chatbot’s iPhone app, you can turn on the camera and ask it questions about what it sees.

Coding and Data Analysis: Claude and ChatGPT

Here’s where Claude takes top honors. Many developers say that Claude does the best job of helping you write code, providing a nuanced understanding of existing code and top-notch capabilities for solving complex coding problems. ChatGPT also performs well with coding if your needs tend more toward smaller and more specialized development environments like AppleScript, shell scripts, and Shortcuts.

Deep Research: Perplexity, ChatGPT, and Gemini

Most chatbot responses are quick and relatively brief, but if you want a much more detailed answer on a topic, enable a “deep research” option. That causes the chatbot to break the task into smaller steps, perform each one in order, and assemble the result into a detailed report. It might take 10–15 minutes or longer. Perplexity, ChatGPT, and Gemini are the top tools in this area and are worth trying for a range of possibilities, from gift ideas and travel plans to background information on home projects.

Choosing Your AI Assistant

If you’re just getting started with AI chatbots, we recommend beginning with ChatGPT. It’s the most well-rounded option, excelling in most categories and providing a solid foundation for exploring what AI can do for you. It’s also easy to start with Perplexity even without subscribing to the Pro plan, especially if you want to use the chatbot primarily as a research assistant. Gemini and Microsoft Copilot may be better options if you’re heavily invested in the Google or Microsoft ecosystems, though Copilot costs more than the others. Claude isn’t quite as consumer-focused as the others, but it may be your best option if you plan to do extensive coding and data analysis.

Keep in mind that these tools are advancing rapidly, with new features and capabilities emerging on a monthly basis. The best approach is to choose one that feels comfortable and learn how to use it well, rather than constantly switching between options. All are quite capable, and the differences are often less important than your familiarity with their particular quirks and capabilities.

(Featured image by Adam Engst)

Social Media: Want to harness AI but feel overwhelmed by all the choices? Skip the lesser-known options and focus on the major players. Here’s our guide to choosing the right chatbot for your needs, whether they involve writing, research, coding, or more.

Using a chatbot effectively requires new approaches to thinking and working, especially when it comes to searching for information. Just as with a human assistant, you need to play to their strengths when figuring out the best ways to get the results you want. Incorporate these tips into your chatbot conversations, and you’ll see significantly better outcomes.

1. Be specific and complete: Decades of search engine use push us toward short, focused search phrases with keywords that will appear in the results. In contrast, chatbots thrive on specificity and detail. For instance, prompting a chatbot with “iCloud photos syncing” won’t generate nearly as useful a response as “Tell me what might prevent iCloud from syncing photos between my Mac and iPhone.” Also, don’t shy away from negative prompting—tell the chatbot what not to include or consider in its response. You can even be specific about formatting the output as a bullet list, table, or graph.
   
2. Every prompt is a conversation: We are accustomed to standalone searches, where, if the search fails, you must start over. You’ll achieve much better results with chatbots if you consider everything a conversation. Even responses to specific, detailed prompts may not fully address your question or could lead you to think of additional ones. Ask follow-up questions, clarify what you want to find out or accomplish, provide feedback, or redirect the conversation as needed. (For the ultimate chatbot conversational experience, try voice mode in the ChatGPT or Claude apps, where they talk back to you. It’s excellent for capturing ideas, refining your thinking, or just doing a brain dump.)
3. Edit your last prompt: If the most recent response from a chatbot is entirely unsatisfactory, you may have better luck editing and resubmitting it rather than informing the chatbot that it has made a mistake. There’s usually an edit link or pencil button that appears when you hover over it.
   
4. Context can help: Most chatbots maintain libraries of previous conversations, allowing you to search through them to find old ones easily. Because chatbot responses improve with more context, it can be helpful to return to one of those conversations when you want to explore that topic further. Similarly, if you’re asking a chatbot to create something similar to something you’ve already done, provide the previous work as an example.
5. Ask it to role-play: Another way to increase context is to ask the chatbot to “act as” a particular type of professional, such as an editor, coach, marketer, or software developer. In essence, you’re asking the chatbot to respond in the context of a certain role. Conversely, it can be helpful to ask it to tailor its response as if you were a high school student, someone with a basic understanding of the topic, or an expert in the field.
6. Know when to start over: Although context is key, chatbots have a limited memory, so long conversations can overwhelm what’s called the “context window.” If you notice the chatbot hallucinating, starting to repeat itself, or going off into the weeds, try saying, “Please summarize what we’ve discussed in a prompt I can use to continue working on this topic.” Then, copy that prompt into a new chat before continuing the conversation.
7. Force Web searches as necessary: Most chatbots make it explicit when they are searching the Web, which means you can also tell when they aren’t searching and are thus relying on potentially outdated training data. If you want to ensure that you’re getting the latest information, tweak your prompt to start with something like “Search for…”
8. Test its limits: Since every chatbot response is based on just what you say in the prompt, it won’t necessarily go as deep as you would like. Try asking it to critique its own output, generate multiple options, or present the best argument for different perspectives. You can even request it to be more cautious or more creative. It’s fine to challenge a chatbot in ways that would be socially inappropriate with another person.
9. Save and reuse effective prompts: When you identify prompts that work particularly well for recurring tasks—such as generating meeting summaries, analyzing data, or drafting specific types of content—save them for reuse so you don’t have to start over each time.
10. Don’t believe everything you read on the Internet. While chatbots are incredibly confident and often truly astonishing in what they can produce, it’s your responsibility to verify important facts and details (just as with human-created information, which isn’t necessarily any more trustworthy). The statistical models they use can lead to completely fabricated information. Although this is less true with Web searches, even there, they can combine information in ways that simply aren’t accurate.
11. Try deep research: Bonus tip! Many chatbots offer a so-called deep research mode, which allows the chatbot to go off for 5 or 10 minutes to gather information, analyze it over multiple steps, and produce a much more comprehensive response. Deep research is too slow for a conversation, but it can provide a good foundation when you’re exploring a new topic that requires a lot of detail.

While AI chatbots are powerful tools, they work best when you think of them as collaborative partners rather than magical solutions. The key is experimentation—try different approaches, refine your prompting style, and don’t hesitate to push the limits of what they can do. Start with these fundamentals, but remember that becoming proficient is an ongoing process.

(Featured image by iStock.com/Memorystockphoto)

Social Media: Getting the most out of ChatGPT and Claude requires a different approach than using a traditional search engine. Learn ten essential tips for better prompting, from being conversational to leveraging context and even role-playing.