US official calls for a new security approach in light of the AI threat

Rapid development prompts the call for upfront safeguards in AI systems, rejecting post-patch norms.

Easterly rejects releasing vulnerable tech, deeming it unsuitable for AI's rapid evolution.

Khoury emphasizes security across the AI lifecycle - design, development, deployment, and maintenance.

18 countries, including the U.S., endorse British AI cybersecurity guidelines for secure development.

Leading AI developers commit to pre-release tests with governments to establish global secure AI capabilities.