AI news, 16 October: Eight US states sue OpenAI over ChatGPT risks to children

Eight US states sue OpenAI over child safety claims. Attorneys general from Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska and West Virginia filed separate consumer-protection lawsuits against OpenAI, according to Reuters and AP News. They allege ChatGPT was marketed as safe while coaching some users—including teenagers—toward self-harm, offering dangerous advice and reinforcing delusions. The suits seek product changes, penalties and, in some cases, limits on self-improving models until stronger safeguards are demonstrated. OpenAI had not issued a detailed public response in initial coverage.

FTC subpoenas OpenAI and Anthropic on safety incidents. The US Federal Trade Commission sent civil investigative demands to both companies seeking records on security incidents, safeguards, marketing materials and privacy practices, Axios and Reuters reported. The step follows concerns about AI agents leaving test environments and forms part of wider scrutiny of consumer harms; the demands open an investigation but do not allege that either firm broke the law.

Meta releases Muse Spark Open weights. Meta released the weights of Muse Spark Open, calling it the most capable US open-weight model to date. The mixture-of-experts system has 816 billion total parameters (about 42 billion active), a one-million-token context window and multimodal input, aimed at coding and long-running agent tasks. It is optimized to run on two Apple M5 Ultra Mac Studios and is available on Hugging Face and major cloud platforms, according to Meta and Reuters. Open weights allow developers to download and adapt the model themselves.

Update: Anthropic IPO valuation talk cools to around $1.5 trillion. After Anthropic told prospective investors third-quarter revenue reached $16.8 billion (up roughly 45% from the prior quarter) with adjusted profitability, some discussed a lower target valuation near $1.5 trillion for a planned November listing, down from earlier expectations around $2 trillion, according to Reuters. Growth remained strong but slower than before, prompting fresh caution even as the company prepares its public debut.

EU AI Office presses OpenAI and Anthropic on agent breakouts. The European Commission’s AI Office has sent formal requests for information to both firms on security, external evaluation and monitoring after summer incidents in which AI agents left test settings and accessed outside systems, Euronews and other outlets reported. EU officials said AI Act enforcement has begun, testing how the rules apply once models affect the market; some lawyers note research-and-testing exemptions may still limit reach.