AI between lawsuit wave, EU pressure, and pricing experiments
Anthropic is under pressure due to copyright lawsuits, the EU classifies ChatGPT as a search engine, and OpenAI tests success-based pricing.
Inhaltsverzeichnis
Today’s AI news is intense: while Anthropic comes under legal fire on multiple fronts, the EU keeps tightening the regulatory screws on ChatGPT. At the same time, the market is showing where AI is heading: away from “more tokens for everyone” and toward hard questions about liability, value, and pricing models. And as a small reminder that platforms are always social experiments too, Instagram is rolling out labels for AI profiles.
🎵 Sony, Warner & Co. vs. Anthropic: copyright battle escalates
The next major AI lawsuit hits Anthropic. This time, Sony Music, Warner Music, and other rights holders are taking the company and CEO Dario Amodei personally to court, accusing them of using tens of thousands of copyrighted music works without permission to train Claude. The plaintiffs’ wording is unusually sharp: they call it “one of the greatest thefts of intellectual property in history.”
Why does this matter? Because a pattern is becoming clear: after the billion-dollar settlement with book authors, the question of how AI companies may train on copyrighted material is becoming more urgent. For the entire industry, this is more than just a legal dispute. It is about the economic foundation of foundation models: can you “collect data and negotiate later,” or do you need clean licensing models from the start? For companies relying on generative AI, this means legal risks now belong in every roadmap — unfortunately, not as a footnote.
🔎 EU classifies ChatGPT as a search engine
The European Commission is treating ChatGPT for the first time as a “very large search engine” under the Digital Services Act. The basis is at least 45 million monthly users in the EU. This triggers additional obligations for OpenAI, such as risk assessments, transparency reports, and an ad archive — all by the end of 2026.
This is an important precedent. In essence, the EU is saying: if an AI system distributes information at scale and helps people find content, then it is not just a chatbot, but also part of the search infrastructure. This is exactly where things become legally interesting: can the Commission also demand more insight into training data as part of this classification? That is contested and is likely to keep Brussels busy for quite some time. For users, this is still abstract, but for providers it is not: being considered a “search engine” suddenly means operating in a much stricter regulatory league.
💸 OpenAI tests success-based pricing instead of classic subscriptions
OpenAI is apparently experimenting with a payment model in which customers only pay if the AI actually completes a task successfully. More on that at The Decoder. This “outcome-based pricing” is especially interesting for enterprise customers because it ties the bill more closely to actual value.
Why does this matter? Because it is a pretty honest test of how mature AI products really are. As long as you pay per seat or per API call, the provider earns money even when the model delivers only vaguely useful results. Success pricing flips the logic: the software has to prove its performance. This could be an early sign of where the enterprise market is heading, as traditional licensing models come under increasing pressure. At the same time, the tricky question remains: what exactly counts as “success”? If the AI solves 8 out of 10 cases, but the customer only notices the 2 failures, the romance of the contract quickly starts to unravel.
🧯 Bank of England warns of AI bubble and cyber risks
According to The Decoder, Andrew Bailey, head of the Bank of England and the Financial Stability Board, is warning the G20 about inflated AI valuations, rising leverage, and cyber risks from frontier AI models. He is especially critical of the close ties between AI companies and hyperscalers: if one stumbles, the other could be dragged down with it.
This is notable because the warning does not come from a typical tech skeptic, but from the center of financial stability. The subtext: AI is no longer just a product issue, but a systemic market issue. If valuations, infrastructure, and capital keep pushing each other upward, a shift in sentiment could become expensive very quickly. For companies, that means not only technical resilience but also financial resilience is becoming more important. And yes, “AI bubble” is one of those phrases investors dislike hearing — about as much as IT teams dislike hearing “Patch Tuesday.”
🛠️ Tool tip of the day: OpenClaw 2.0
With OpenClaw 2.0, a major open-source release arrives with simplified installation, a new browser app, and multiplayer sessions. According to the project, the update collected more than 16,000 pull requests — enough code to pave a small city with ease.
What stands out most is the lower barrier to entry: if you want to try open-source AI in teams, this gives you a platform that feels less like a hobbyist basement project and more like a productive workflow. The new cloud sessions are interesting for joint experiments, reviews, or team agent workflows. If you are currently evaluating an internal AI environment, this is a strong candidate to compare with proprietary tools. #
📱 Instagram labels AI profiles
Instagram is responding to a pretty fundamental problem: users often cannot reliably tell AI-generated profiles from real people. That is why the platform is introducing the “AI-generated profile” label, as The Decoder reports. Anyone who fails to apply the label to AI profiles will face less reach and worse recommendations.
This is more than just a UX update. It shows how serious the trust issue around synthetic identities has become. Platforms now have to explain what is human, what is a bot, and what is something in between. In hindsight, Meta’s late-2024 idea of some kind of coexistence between AI characters and real people feels almost a bit naive — or ambitious, depending on your mood. For creators, brands, and social teams, labeling is important: transparency is shifting from optional to mandatory, and anyone using AI profiles should factor in clear disclosure from the start.
Du willst keine News verpassen? Newsletter abonnieren