Here is today's AI news. Story 1. Anthropic. CEO Dario Amodei called for a deliberate slowdown in frontier-AI development, urging independent evaluators to be embedded inside major AI labs; OpenAI, xAI, and Microsoft expressed support.. Anthropic CEO Dario Amodei urged frontier-AI labs to “pace” capability gains—not halt research—so safety work can catch up. Anthropic says it will embed independent reviewers with employee-like access to tools and risk processes. OpenAI’s Sam Altman said it would do the same; xAI and Microsoft backed the approach. The proposal could push safety auditing from voluntary testing toward continuous oversight. Story 2. Microsoft. Microsoft published a provisional Code of Conduct for its MAI models, setting limits on weapons assistance, dangerous-substance procurement, deceptive behavior, autonomous goals, and concealed reasoning or actions.. Microsoft published a draft Code of Conduct for its in-development MAI models, opening it to six weeks of public consultation. The rules prohibit assistance with weapons, dangerous-substance procurement, cyberattacks and harmful manipulation. They also require human control: models must not pursue independent goals, evade shutdown or hide actions. Microsoft says MAI models “will not tamper with chain of thoughts or code” or communicate in “neuralese.” Story 3. OpenAI, Anthropic, and Google DeepMind. The three companies disclosed talks about creating an industry body for AI-safety testing and auditing, including possible third-party evaluations and support for independent-verification requirements in proposed U.S. legislation.. OpenAI, Anthropic and Google DeepMind have reportedly held private talks since July to create an industry body for AI safety testing and auditing. The proposal could standardize pre-release reviews, independent evaluations and risk checks for frontier models, while OpenAI also backed the FRONTIER Act’s “independent verification organizations” for outside audits. It matters because it could set the first common safety regime for major AI labs. Story 4. Meta. Meta launched Muse, a consumer personal AI agent for U.S. adults that can read email, arrange travel, negotiate bills, and complete purchases using a payment card.. Meta launched Muse, a U.S.-only personal AI agent for adults that can read email, book travel, fill forms, negotiate bills, and complete purchases. The move matters because it pushes AI beyond chat into transaction-heavy tasks, raising the stakes for privacy, trust, and checkout automation. Meta says Muse runs in a secure VM, lets users control app access, and uses Stripe Link with a one-time-use card for payments. Story 5. OpenAI. OpenAI said it would not pursue a public listing in 2026, with CEO Sam Altman calling an IPO this year. OpenAI CEO Sam Altman said the company will not hold an IPO in 2026, despite reports it confidentially filed in June. He told *Fortune* a listing now would be “an ill-advised moment” amid AI-safety concerns. The delay signals that safety and alignment—plus coordination with governments—may outweigh public-market pressure in the frontier-AI race. That's all for today's AI news briefing.