Demis Hassabis Calls for U.S.-Led AI Watchdog Before Model Release

The Neuron · 1 day ago

Google DeepMind CEO Demis Hassabis published a manifesto proposing a U.S. national AI standards body to evaluate advanced models for security and safety risks before public release. The proposed independent board, modeled on financial industry regulator FINRA, would be funded by AI companies and become operational by the end of 2024. If adopted, such a watchdog would require developers to submit models for testing on cyber, biological, and deception capabilities before deployment.