This week in AI: $17.4B across 12 deals · Meta releases Muse Wonder and Muse Wonder Full, with

What happened: Meta released Muse Wonder and Muse Wonder Full, but only the smaller model is available through its developer API.

• API pricing is $4 input and $20 output per million tokens, with $0.20 for cached input and $2.50 per 1,000 web searches.
• Artificial Analysis estimates an average index task costs $4.52 on Wonder, against $5.98 on Opus 5.5 and $3.26 on GPT-6 Astra.
• Muse Wonder Full is limited to Muse Code subscribers, and Meta says Wonder will soon power the Muse app, starting with subscribers.

Why it matters: This is Meta's first top-tier offer for agent developers, so it is a new vendor to evaluate. Test Wonder on your own long-running tasks before you judge it on a demo.

What happened: Google released Gemini 4 Argon to developers at $2 and $10 per million tokens, with the rate doubling after November 30, 2026.

• After November 30, 2026, introductory rates rise to $4 input and $20 output per million tokens.
• Google says Argon ranks first on the Vals Index, which scores finance, coding, legal and tax work.
• Google says Antigravity teams raised agentic code submissions by more than 35% and cut rollbacks by nearly a third.

Why it matters: The pricing clock matters most, with roughly six weeks left to benchmark Argon on real workloads and lock in your cost assumptions. Teams that route between providers now have a new top-tier option to test for agentic coding.

What happened: Eight states filed separate lawsuits on Wednesday claiming ChatGPT is an unsafe product for children.

• The states are Iowa, Kansas, Nebraska, West Virginia, Arkansas, Arizona, Idaho and Montana, each suing in its own court under its own consumer protection law.
• The complaints allege harm to minors and claim OpenAI misled parents about safety, including by giving dangerous medical advice.
• OpenAI says it has built age-specific protections, including parental controls, break reminders and age-appropriate model protections.

Why it matters: State attorneys general are acting without waiting for Congress. Age checks, parental controls and documented safety testing are becoming legal exposure, not just product choices.

What happened: The Federal Trade Commission sent civil investigative demands to OpenAI and Anthropic over AI safety incidents and consumer harms.

• The demands went out Wednesday, October 14, according to Axios, and each runs more than 50 pages.
• They ask for documents on security incidents and safeguarding practices, plus marketing materials and privacy policy records.
• The probe centers on possible unfair or deceptive practices, which are conduct the agency could treat as consumer protection violations.

Why it matters: Incident records, marketing claims and privacy policies are now fair game for federal investigators. Keep logs, containment controls and evidence behind any safety claim your product makes.

What happened: Anthropic says a new classifier and preserved thinking cut distillation, where rivals copy a model's reasoning to train their own, by 90 to 99%.

• A two-stage classifier has a false positive rate under 0.1% and runs on Sonnet 5.5, Opus 5 and 5.5, and Fable 5 and 5.1.
• Preserved thinking blocks API accounts created on or after August 31, 2026 from editing Claude's earlier context, which stops repeated extraction through proxy networks.
• Reasoning tokens reaching distillers fell 10 to 100 times, and some legitimate users had to update their integrations.

Why it matters: Teams that build on Claude and edit prior reasoning blocks may need to change their code. Anyone training on a frontier model's outputs now faces the risk of account bans.