A Silicon Valley congressman is pushing new AI rules that would hit the biggest AI companies — bad news for OpenAI and Anthropic, good news for safety advocates who want outside referees.

• Ro Khanna is introducing a bill to ban AI that can rewrite and retrain itself without human control.
• The bill would also put independent auditors inside AI companies and create an FDA-style AI safety agency.
• Tech billionaires cannot be trusted to write their own rules, Khanna argues, calling Trump's voluntary compact with AI firms absurd.
• He wants the new agency staffed by safety researchers with no financial stake — not industry insiders, and not Bill Gates or Dario Amodei, both tainted by Epstein links.
• He also backs banning data centers in neighborhoods and farmland, and says hospitals are now using AI to pad billing codes and squeeze insurers and Medicare.

Outlook: The bill faces an uphill fight in a Washington that currently favors letting AI companies police themselves, but data centers and AI jobs are becoming live election issues.