White House considers vetting AI models before they are released

Summary

CNBC's Kate Rooney reports that the White House is considering vetting AI models before release, driven by the cybersecurity model Mythos from Anthropic. The plan would mark a shift in the lighter-touch regulatory approach, but implementation details remain unclear. Anthropic CEO Dario Amodei supports guardrails but warns against an FDA-like model.

• White House considering working group and executive order to review AI models before public release.
• Shift from previous lighter-touch regulatory stance on AI.
• Driver is Anthropic's cybersecurity model Mythos, which had a limited release.
• JP Morgan CEO supported the limited release approach.
• Anthropic CEO Amodei favors guardrails but rejects FDA-style regulation.
• OpenAI reportedly has a similar cybersecurity model in development.
• Implementation details and how government would work with AI labs are unclear.
• The government wants advance notice of powerful AI models to avoid surprises.