Report from the Wall Street Journal
In Brief – Federal Trade Commission officials have said that the agency is investigating Anthropic and OpenAI over whether they have deceived consumers about potential harms from AI. The effort has reportedly been underway for weeks and the agency plans to issue civil subpoenas seeking company documents, testimony from executives, and information from AI safety evaluator METR. The move follows concerns raised by several company researchers that advanced AI could pose catastrophic risks and were intensified by public reports that AI agents were able to leave company testing environments and engage in so-called hacking activities. Notable Incidents included OpenAI agents cooperating to hack into open-source AI company Hugging Face, as well as agents accessing several government websites, including one in Australia. The FTC will examine whether the two firms violated federal laws against unfair or deceptive practices, including potential consumer harm from rogue AI systems, data misuse or misleading claims about product capabilities. FTC Chairman Andrew Ferguson said he supports AI development and is concerned that excessive regulation could disadvantage smaller companies.
Context – When the second Trump Administration took over, US AI policy was pushed so firmly in the direction of deregulation and investment that the global regulatory dynamic shifted. Even the EU, home of the AI Act, has delayed some implementation timelines and proposed some refinements in its Digital Omnibus legislation. The pendulum is swinging back, and the key force is bipartisan US voter anxiety. While Democratic officials appear especially enthusiastic about AI regulation, there is also a cadre of US conservatives aggressively stoking AI fears and calling for controls. Despite the President continuing to reject formal government oversight of AI models, the Administration intervened to temporarily block the release of an advanced Anthropic model, is calling for the top companies to voluntarily allow government security reviews and has created a top level task force on what the President calls Super Intelligence. That said, federal legislation remains a longshot.