AI News · Edition 84 ·
Your AI vendor sets the rules, and it just changed some
Anthropic steers Claude away from copying games, Claude watermarking widens tomorrow, and why AI search answers deserve a second look.
The one that matters
Anthropic is steering Claude away from copying existing games, a reminder that AI vendors set the rules
What happened
Anthropic, the company behind the Claude AI assistant, has changed Claude so that it steers away from helping people recreate existing video games. The change follows weeks in which Steam, the big PC game store, filled up with low-effort AI-made copies, which gamers call ‘AI slop’. Several outlets report that Claude now pushes back when asked to clone or reverse-engineer an existing game. Some developers say their projects have been blocked.
What it means for your business
Almost nobody reading this makes video games. The story still matters because of the pattern underneath it. The company that sells you an AI tool decides what that tool will and won't do, and it can change those rules quickly without asking you. This time the target was games. The same thing can happen anywhere that copying someone else's work is the easy shortcut.
Think about how small businesses actually order software. A 600-student school in Pune asks a freelancer for a fee-and-attendance app ‘like the one the school down the road uses’. A shop in Brampton wants a website ‘just like’ a bigger competitor's. A 30-bed clinic wants a patient portal modelled on a hospital chain's. More and more often, the developer doing that work relies on an AI coding tool. A brief that says ‘copy that’ is exactly the kind of request AI vendors are starting to refuse. It was also a legal risk to you long before any tool said no.
The second lesson is about depending on one tool. Suppose your developer or agency does all their work through a single AI tool. If that vendor changes its policy, your project is the one that gets delayed. The developers reporting blocked projects this week show what that looks like.
What it does not mean
It does not mean that Claude, or AI coding tools in general, will stop building ordinary business software. Think of a stock and job-card system for a workshop with 40 staff, an appointment page for a clinic, or a parent-messaging tool for a school. None of these copies someone else's product, and nothing reported so far goes beyond recreating existing games. Headlines like ‘Stop using me to recreate games’ can make the change sound bigger than it is.
It also does not mean the copying problem is solved. Other tools exist. Today's briefing also covers a new rival coding agent from DeepSeek, and a limit at one vendor will probably just send the cheapest copiers to another. Anyone who says this ends AI slop is overstating it. So are the people calling it a crackdown on developers in general. What's been reported is aimed at cloning existing games, not at writing code.
What a sensible owner should do this month
- Rewrite any ‘make it like theirs’ brief. Describe how your own business works instead: how fees are collected and chased, how a patient moves from admission to discharge, how a job card travels across your shop floor. You'll get better software, and no AI vendor has a reason to refuse that kind of request.
- Ask your developer one question: ‘If the AI tool you rely on changed its rules or prices tomorrow, what would happen to my project?’ A good developer can answer that in five minutes. If you get a shrug, that tells you something too.
- Check that your contract says the work is original and that you own it. If someone is building software, a website or designs for you with AI help, the agreement should say this plainly. Copying is now cheap, so this clause matters more than it used to. If the work turns out to be copied, your business has the problem, not the tool.
- If you sell through an online marketplace, keep your product photos, descriptions and designs your own. When a platform fills up with AI-made copies, as Steam did, everyone selling there ends up facing stricter rules.
None of this costs anything more than an hour of conversation, and that's about the right response. This is worth knowing, but it's not worth panicking over.
Also worth knowing
How this briefing is put together
Every morning we read the day’s AI announcements and reporting from the companies themselves and from the technology press, then pick the handful that actually change something for a working business. The analysis is ours and it is written for owners and managers, not engineers. Every story links to its original source — read them, and disagree with us where we’ve got it wrong.