Insiders who built today's AI systems are quitting and warning the technology could wipe out humans within a decade — bad news for anyone hoping the industry polices itself.

• A former OpenAI researcher says companies are racing to hand AI control of everything, including the military, before anyone knows if it can be trusted.
• An Anthropic researcher resigned saying neither OpenAI nor Anthropic is acting responsibly and both are gambling with human lives.
• Anthropic's own alignment lead put the odds of AI killing everyone at more than 10% in the next decade, and admits there is no plan to fix it.
• The core fear is AI that rewrites and improves itself faster than people can follow, leaving humans out of the loop.
• Washington is moving the other way: a $125 million AI super PAC is pushing a federal framework that would cancel out AI laws passed by 38 states.

Outlook: Expect louder insider warnings and congressional hearings, but no real safeguards while the US frames AI as a race it must win against China.