From the article ...
Anthropic has drawn two red lines: no mass surveillance of Americans and no fully autonomous weapons ... OpenAI, Google and xAI—have agreed to loosen safeguards ... Pentagon has demanded that AI be available for “all lawful purposes.”