Sunday’s AI tape is a recap of a week, not a ship date. Reuters published a September 19 feature that strings together ten days of agent-breach notes, researcher exits, and CEO statements at OpenAI, Anthropic, Google DeepMind, Microsoft, and xAI after Dario Amodei’s September 12 essay.

The piece starts the clock at Jacob Coxon leaving Anthropic on September 8 and runs through public slowdown language from Amodei, Altman, Hassabis, and Musk. It does not add a new model name. Separate Sunday write-ups are still repeating Friday’s Reuters exclusive that Anthropic is considering a release before IPO. That remains a source story. See Saturday’s brief.

Key takeaways

What was actually committed

Independent evaluators with staff-like access was Altman’s public pledge last weekend. A standards-body conversation among OpenAI, Anthropic, and Google has been acknowledged by OpenAI policy staff. Neither item is a freeze on training. Sunday adds the long feature, not a filing.