“You f*cking idiot, I asked for…” might seem harmless if you’re saying it to an AI chatbot, and a lot of people are doing it. But if you’re one of them and your LLM of choice is Anthropic’s Claude, you might want to start keeping your frustration-driven interactions in check.

Anthropic recently updated its fine print regarding how users are allowed to interact with Claude. The changes come into effect on November 12th, 2026, and they’re part of a wider update to Anthropic’s terms of service.

According to this 2026 Usage Policy Update post, the goal is to address “extreme cases where users repeatedly act cruelly toward our models, with no discernible purpose.” The rule doesn’t apply to “common versions of user frustration, pushback, dark creative themes, or model testing and research.” (It’s unclear where the line sits between “frustration” and “cruel behavior.”)

The full relevant text:

Addressing abusive behavior toward our models

We’ve added a prohibition on sustained and needless abusive or cruel behavior toward our models. The policy update is meant to apply only in extreme cases, where users repeatedly act cruelly toward our models, with no discernible purpose. It does not apply to common versions of user frustration, pushback, dark creative themes, or model testing and research.

This addition aligns with a step we’ve already taken, allowing Claude models to end rare conversations with persistently abusive users on Claude.ai and Claude Code. Such abuse is the main focus of this update; Claude’s ability to end these interactions will remain the primary enforcement mechanism.

What happens if Claude is offended by your behavior? The main consequence is a terminated chat. Anthropic has already given its models the ability to end chats with users who persist in abusive behaviors on Claude.ai and Claude Code—this change enshrines that as an explicit prohibition in the company’s terms of use.

The update also includes clarifications on: political content and election interference, surveillance and law enforcement, weapons technology, and high-risk use cases involving health, legal rights, and finances.

Yesterday, Microsoft dropped the amount of OneDrive cloud storage included with Microsoft 365 Family, Premium, and Pro. The change only affects new customers right now, but existing customers will get hit with the reduction at their first subscription renewal on or after May 2nd, 2027.

Prior to this change, Microsoft 365 Family, Premium, and Pro provided 1TB of OneDrive storage per user, with all plans able to be shared with up to 5 other users (for a total of 6TB per account). After this change, both Family and Premium accounts are limited to a 2TB storage pool that’s shared by all users. Pro accounts get 5TB of cloud storage to share.

It’s a mixed bag. Solo users who were previously capped at 1TB now get twice the storage to play with, and anyone who was sharing with only one other person isn’t facing a huge difference. But if your account was split between three or more users, it’s a big downgrade.

What stings the most is that Microsoft isn’t adjusting the pricing on any of the plans. Microsoft 365 Family remains $12.99/month or $129.99/year, while Premium stays $19.99/month or $199.99/year. Meanwhile, Microsoft 365 Pro is a whopping $99.99/month with no annual option.

On this support page regarding the change, Microsoft makes it clear that it’s one shared storage amount: “The primary subscriber and up to five additional members draw from the same shared storage pool.” However, that doesn’t mean access is shared between all users: “While storage quotas are shared, files remain private.”

And you can expand the size of the storage pool with add-on purchases. It doesn’t sound like individual users can boost their own individual caps, but rather only contribute to the shared cap: “Storage add-ons purchased by subscription members are automatically added to the shared pool. Eligible family members’ free storage grants also contribute to the shared pool.”

The non-shareable plans remain unchanged: Microsoft 365 Basic gets 100GB of storage for $1.99/month or $19.99/year, while Microsoft 365 Personal gets 1TB of storage for $9.99/month or $99.99/year.

From treasured family photos to essential client footage, losing files can be devastating. Whether it’s down to a corrupted SD card, sudden power loss, or accidental deletion, recovering as much of that data as possible is critical. Wondershare Recoverit V15 is built to cover thousands of loss scenarios, across over a million different storage devices, giving you the best chance of salvaging that data.

With over 20 years of data recovery experience, Wondershare’s services combine robust and effective file recovery for photos, videos, and other data, with a streamlined process that helps you dig through even large file systems to recover and organize the most important files.

No SD Card Left Behind

For creators with hundreds or thousands of photos and videos on SD cards, an accidental deletion or sudden corruption event can feel catastrophic. That’s where Recoverit V15’s Profession SD Card Recovery can step in and help prevent disaster. It can identify the device and card information to use a bespoke scanning strategy to maximize file recovery and minimize data loss. You can track the recovery process the whole way, too, letting you preview photos and videos in real time, so you begin file recovery as soon as you confirm the footage is viable, without having to wait for the scan to finish. It has full support for the most popular video and image resolutions, too, including HDR and high-bitrate video.

After recovery it can organize the recovered files based on camera directory structures and the specific relationships between different files, making it easier to find the exact files you need to get back.

This works across other external storage solutions, too. If you’ve lost data on a corrupted flash drive, memory stick, or CF card, Recoverit V15’s Profession SD Card recovery tool is an excellent solution to the problem.

AI Advantage: Find Exactly What You Need

When recovering large quantities of lost media files, finding the most important photos and footage with traditional recovery tools isn’t easy; Recoverit V15’s smart search helps you narrow the results. It uses large language models to generate file tags and facilitate natural language search, helping you to find what you need.

Recoverit V15 groups scanned files for more focused recovery, linking related RAW and JPEG files, pairing videos with associated subtitles, and making it clear which files are ready to recover, and which ones may need further repair. In-built recovery quality evaluation ensures users have a clear picture of the state of the data. Quality checks show the condition of the scanned files, while automated repair tools attempt to fix any damage.

A Tool For Everyone

Recoverit V15 is compatible with Windows and macOS systems, allowing file and partition recovery on either operating system, with support for Apple Silicon, including the newer M5-based Macs and Mac Neo. Whether you’re running an older version of these operating systems or the latest Insider previews, Recoverit V15 is built to offer fast, effective file recovery with the same streamlined interface.

Give Wondershare Recoverit V15 a go today to help recover your lost files.

Why in the world would you ever give ChatGPT access to your email? It’s a reasonable question, given everything we’re hearing about rogue AI agents, not to mention the privacy implications of unleashing AI in your inbox.

Then again, one of the best ways to appreciate the power of a personal AI assistant is to let it triage your unruly inbox, culling the spam and the marketing blasts while highlighting the messages that truly matter.

An agent with access to your messages can also give you a heads-up about impending meetings, warn you that your building’s elevator is out of service (that happened to me earlier this week), keep you current on that Jira ticket you filed, and perform other powerful workflows (like the one in this week’s prompt of the week, below).

To get access to your email, AI services like ChatGPT, Claude, Gemini, and Meta’s Muse agent need tools, and among the most common AI tools for email are those for Gmail. These tools allow AI models and agents to perform certain Gmail tasks, like “read message,” “add label,” “draft message,” or even — gulp — “send message.” 

They also come with sets of rules and permissions that (generally speaking) you can customize, allowing some actions to happen autonomously while others require a user’s approval, or are blocked altogether. Tweak these rules the right way, and you can minimize the chance of your agent, say, sending an email on its own when it shouldn’t. 

The key word, of course, is “minimize.” AI agents like ChatGPT Dots, Claude Cowork, Gemini Spark, and Meta Muse can act in unpredictable ways, and while you can prevent the agents from scooping up your Gmail messages for AI training data, sensitive details from your Gmail conversations could wind up in your AI chat threads, where they could be stored as memories.

So if you’re uncomfortable taking those risks — and hey, I get it — my advice is to leave the Gmail tools alone.

If you do choose to move ahead, know that enabling an AI’s Gmail integrations is just the first step. (ChatGPT has a Gmail “plugin,” while Claude and Muse call them “connectors,” and Gemini folds its Gmail integration under its “Personal Intelligence” umbrella.) Dig deeper, and you’ll find more granular controls for those Gmail tools, where you’ll be able to do things like set guardrails for when an agent can or can’t send email on its own. 

For Claude, click Settings > Connectors, then click on Gmail. From there, you’ll find a wide range of Gmail rules, from allowing Claude to peruse received and drafted messages to whether it can add labels to messages or even send them on its own. You can set entire groups of read-only and write/delete rules to “Always Allow,” “Needs Approval,” or “Blocked,” or you can set specific rules for each action. 

ChatGPT’s Gmail plugin rules (check Settings > Plugins) aren’t nearly as granular as Claude’s, giving you just four total options: “Always ask,” “Allow read actions,” “Allow low-risk actions,” or “Allow all actions.” The second setting — “Allow low-risk actions” — is the default, but personally I’d stick with “Allow read actions,” just to be sure ChatGPT asks before sending email on its own. 

Gemini is even more vague about the rules for its Gmail tool, essentially limiting you to turning its Gmail integration either on or off. According to a Google support page, Spark is “designed to ask for your review” before “sending communications,” which means it should stop and ask before sending a Gmail message on its own. But if Gemini’s lack of specific permission controls for Spark gives you pause (and that’s perfectly understandable), you might want to switch its Gmail connector off.

Finally, Meta Muse keeps its tool settings under Settings > Permissions > Connectors. Select the Gmail option, then start tweaking. As with Claude, Muse’s permission settings for Gmail allow you to fine-tune whether it can do things like access and search your email, draft messages, and send emails on its own, with your approval, or never. 

Aside from general Gmail permissions, you’ll want to block the use of your Gmail for AI model training. The best way is to ensure you have the overall “help improve our models” setting for ChatGPT, Claude, and Meta Muse turned off (check under Settings > Data Controls or Privacy). Google, for its part, says it won’t train its models “directly” on your Gmail messages, but may train on “limited info,” such as “specific prompts in Gemini or AI Mode and the model’s responses.”

Now that you’ve hardened permissions and checked your AI model-training settings, what can you do with your new Gmail AI superpowers? Check the prompt below for one example, and I’ll have more in the weeks ahead.

More in AI this week

• Google will now let anyone try its SynthID detector, a tool that lets you check images, videos, and audio files for AI watermarks. Just keep in mind that the detector can’t prove whether a suspicious-looking viral video is real. [PCWorld]
• ChatGPT is about to get its own digital wallet, which could allow ChatGPT Dot agents to make web purchases on its own. [PCWorld]
• Common Sense Media is sounding the alarm over ChatGPT for Teens, which “could give parents false confidence in guardrails and safety alerts that frequently don’t work.” [The Verge]
• Anthropic’s just filled out its 5.5 family of Claude models with the fast and cheap Haiku 5.5, which joins the mid-range Sonnet 5.5 and the top-end Opus 5.5. I plan on giving the new Haiku a test drive in the next few days, so I’ll report back on what it can and can’t do. [VentureBeat]
• AI is crushing it when it comes to solving math’s toughest riddles. But while their solutions check out, their reasoning is often indecipherable to human mathematicians, with some grimly predicting the “end of math as we know it.” [Quanta Magazine]

Prompt of the week: The “find my forgotten accounts” prompt

So, you took a deep breath and connected your Gmail to AI. Here’s an agentic “outcome”-oriented prompt to put that new ability to the test.

With this prompt, you’ll be directing an AI agent to scour your Gmail for messages about long-lost accounts you previously signed up for, with the AI doing a read-only search for “verify your email” and “confirm your account” messages.

When it’s done, you’ll get a spreadsheet detailing all the accounts the agent found, with the most dormant at the top, along with links (when available) for deleting the forgotten accounts you don’t want anymore.

That’s all for now!

Thanks for reading the latest issue of Prompt Mode.

Want more next week? Don’t forget to sign up to start receiving this newsletter in your inbox.

I don’t know about you, but there’ve been countless times that I go to bed exhausted only for my brain to suddenly decide it’s the perfect time to replay an awkward conversation from five years ago. If that sounds familiar, then I recommend giving the site Fall Asleep a try. It’s a free website with a small collection of exercises aimed at helping you quiet your thoughts so you can drift off to sleep.

Shutting your brain off at night can be easier said than done, especially when your whole life is full of awkward moments. The most successful method I’ve found is called “cognitive shuffling” and Fall Asleep has a really neat tool for this. Cognitive shuffling serves up random, unrelated words and images that you then visualize and focus on in your mind. The idea is to break up the connected, problem-solving thoughts that tend to keep your brain switched on when you should be sleeping.

Cognitive Shuffling works best for me, but Fall Asleep offers a few other tools like a simple guided breathing exercise and a tool called “Brain Dump” that encourages you to jot down your busy thoughts before bed. According to the site, nothing is shared and all data is stored locally on your web browser.

Unlike sleep tracking smart watches that overload you with anxiety-inducing data metrics, Fall Asleep keeps it low-key and that’s for the best. So the next time you’re lying awake at 1 a.m. staring at the ceiling, go ahead and fire up Fall Asleep and see if it helps. 

For more fun and helpful tips, tricks, website and app recommendations, be sure to subscribe to our PCWorld Try This newsletter.