Trump Admin: Voluntary AI Safety Reviews Before Launch

The White House plans to introduce a new executive order proposing voluntary government reviews of advanced AI models. This framework could require companies like OpenAI and Anthropic to share their frontier AI systems with US agencies. The reviews would happen up to 90 days before public launch for cybersecurity and national security evaluations. What's interesting is that the order aims to balance the development of AI technologies with concerns about potential cybersecurity risks and threats to critical infrastructure. The plan calls for AI firms joining the program to give the government access to certain high-risk AI models before they are launched. Agencies like the National Institute of Standards and Technology and the Cybersecurity and Infrastructure Security Agency would create benchmarking standards. The NSA would reportedly make the final decision on which systems are included for review. The bottom line is this move could significantly expand government oversight of advanced artificial intelligence systems before they are publicly released.