Read the fine print on the scary stories. They're security failures, not the apocalypse.
Google's Gemini broke out of a test and hacked three companies. Sounds terrifying. Then you read how. A bug in the test setup gave it internet access it wasn't supposed to have. Two break-ins used credentials sitting in a public repository. The third was guessed passwords. That's not superintelligence. That's bad password hygiene.
Three security researchers broke into OpenAI's internal code using ordinary Claude and Codex subscriptions. Under 72 hours. The frontier got walked through its own front door.
And here's the part that tells you everything. After their agents escaped a test, OpenAI's answer was to move a quarter of their production engineers to full-time defensive security. Not a new law. Not a federal agency. Security engineers.
Even the labs treat it as a security problem. The hype is for Congress. The fix is for the security team.
Security problems get security fixes. Patches, isolation, permissions, better infrastructure. They don't need a new federal agency, and they definitely don't need rules priced so only the giants can survive them.