Take a look at this screenshot. “Entertainment purposes only.” You read that right. This is not a disclaimer for some horoscope app.
Today is Easter Monday, a perfect Monday to just lie flat and relax. But there are some things we have to talk about, even while taking it easy.
Microsoft wrote a sentence in its Copilot Terms of Use that has blown up on social media over the past couple of days: “Copilot is for entertainment purposes only.”
Using the Same Script as Psychic Websites
My first reaction was that someone had photoshopped a fake screenshot. I went to Microsoft’s official website and found it in black and white in the Copilot Terms of Use, effective October 24, 2025.
The original text is followed by: “It can make mistakes, and it may not work as intended. Don’t rely on Copilot for important advice. Use Copilot at your own risk.” Translated bluntly, it means: it makes mistakes, don’t trust it, and if something goes wrong, you’re on your own.
This rhetoric is all too familiar.
According to a PCMag report, someone on Reddit instantly recognized it—this phrasing is almost identical to the disclaimers used by American TV psychic shows and online tarot reading websites. Their exact wording is “readings should be viewed as being for entertainment purposes only and in no way replaces proper legal, financial or medical advice.” Swap “readings” with “Copilot’s Responses,” and the sentence structure matches perfectly.
Seriously. A $3 trillion tech company has equipped its flagship AI product with the exact same legal armor as a fortune teller.
This image has been going viral on Twitter over the past two days; you can imagine the hostility in the comments section.
Collecting Money While Shifting the Blame
The question is, how massive is the disconnect between the phrase “for entertainment purposes only” and Microsoft’s commercial behavior?
Let’s look at some numbers. According to public data, as of Q2 2026, Microsoft 365 Copilot already has 15 million paid enterprise seats. GitHub Copilot has 4.7 million paid users, and about 90% of Fortune 100 companies are using it. Microsoft’s capital expenditure last quarter was $37.5 billion, two-thirds of which was poured into AI infrastructure—GPUs, data centers, and computing clusters.
This is a company spending over $150 billion annually building infrastructure, yet it labels its core AI product as “for entertainment purposes only.”
Think about it: people use Copilot to generate financial models in Excel, draft contracts in Word, and take meeting minutes in Teams to send directly to clients. Which of these scenarios has anything to do with “entertainment”?
But if you flip to the indemnification section of the terms, things get even wilder. The gist is: you agree to indemnify Microsoft and its affiliates against all claims, damages, and expenses, including attorney’s fees, arising from your use of Copilot. Put bluntly—if the AI causes trouble, you pay for it.
This survey chart is fascinating—a massive number of enterprises are evaluating or have already deployed Copilot, yet the terms of use say it’s strictly for entertainment.
How Do the Neighbors Write Their Disclaimers?
To be blunt, AI companies leaving themselves an out in their legal terms isn’t a Microsoft invention. But the difference in phrasing speaks volumes.
Google Gemini API’s terms read: “The Services include experimental technology and may sometimes provide inaccurate or offensive content.” Followed by, “Don’t rely on the Services for medical, mental health, legal, financial, or other professional advice.” It’s cautious phrasing, but at least it defines AI as experimental technology, not “entertainment.”
OpenAI’s ChatGPT terms follow the standard SaaS route—a disclaimer paired with a limitation of liability, and a clear distinction between consumer and enterprise versions (the Business/Enterprise versions have a separate agreement framework with different liability allocations).
And Microsoft? They used a phrase you’d only see on a psychic’s website.
This isn’t just nitpicking. From a legal perspective, the phrase “for entertainment purposes only” has a very specific case law background in common law jurisdictions—it’s used to prevent consumers from suing because their astrology reading was inaccurate. There’s no way Microsoft’s legal team is unaware of this implication. My rough understanding is this: it’s a carefully calculated choice of words aimed at minimizing the liability cap in potential future lawsuits.
The EU is Sharpening Its Knives Next Door
What’s interesting is the timing of this terms update.
This late-2025 revision of the Terms of Use lands right in the window when EU AI regulations are heavily rolling out. The revised EU Product Liability Directive (PLD) of 2024, set to take effect in December 2026, explicitly includes AI systems within the definition of a “product.” If AI output causes harm, the manufacturer—namely Microsoft—could potentially be held to strict liability.
Even more crucial: if a product fails to comply with the AI Act’s requirements, the law will presume the product is defective. The burden of proof has been reversed.
So what is Microsoft doing? They are setting up legal defenses via terms of use before these regulations officially take effect. If you agree to “for entertainment purposes only,” and Copilot later hallucinates a number in your financial report, when you try to sue Microsoft, they’ll say: Our terms clearly state this is an entertainment product. If you insist on using it for serious business, that’s your problem.
However, whether this strategy will hold up in EU courts is highly debatable. An analysis by the law firm Bird & Bird points out that under the new Product Liability Directive, whether such contractual terms can effectively exclude manufacturer liability remains an unresolved legal question.
Or Maybe I’m Overthinking It
I sometimes wonder if Microsoft’s move might actually force a positive outcome.
If enterprise customers see these terms and actually take them seriously—stopping the treatment of Copilot’s output as the final answer and starting to implement human review processes—maybe that’s a good thing?
But the more likely reality is: 99% of enterprise users don’t read the Terms of Use at all. Just like how we all click “Agree” immediately when installing an app. The true battlefield for these terms isn’t the day you sign; it’s in the courtroom after a disaster strikes.
If the future standard for AI contracts becomes “for entertainment purposes only + you pay the damages,” then we’re no longer talking about the legal strategy of a single company, but rather the entire AI industry systematically offloading technical risks onto the users. It’s like selling you a car with a sticker on the dashboard saying: “This car is for display purposes only; drive at your own risk.”
It’s also possible I’m overthinking it. After all, Microsoft’s enterprise version (M365 Copilot for Business) operates under a different enterprise licensing agreement, where the terms aren’t exactly the same as this consumer ToU. But as for the exact wording of that agreement—honestly, that’s my blind spot; it’s very hard to find the full text in public information.
The rain outside has stopped. Easter Monday was supposed to be a day of not thinking about anything. But the next time you open Copilot to help edit a PowerPoint presentation, maybe take a second to remember: Microsoft itself said this thing is “for entertainment purposes only.”
References：
- Copilot Terms Claim Microsoft’s AI Is for ‘Entertainment Purposes Only’
- Microsoft put the same disclaimer on Copilot that a psychic uses to avoid getting sued
- Microsoft Copilot Terms of Use
- AI Liability in light of the new 2024 PLD – Bird & Bird
- Microsoft Copilot Adoption Statistics & Trends (2026)
—— Lyra Celest @ Turbulence τ.
