Breaking News

AI Firms Required to Share Safety Test Results with US Government

The Biden administration is set to enforce a new requirement for major artificial intelligence (AI) system developers to disclose safety test results to the government. The White House AI Council is scheduled to meet to assess progress on the executive order signed by President Joe Biden three months ago. One of the key goals of this order, to be achieved within 90 days, is a mandate under the Defense Production Act, compelling AI companies to share crucial information, including safety tests, with the Commerce Department.

Ben Buchanan, White House special adviser on AI, emphasized the government’s intention to ensure the safety of AI systems before their public release, stating that companies must meet this standard. Although software companies have committed to specific categories for safety tests, there is currently no common standard for these tests. The National Institute of Standards and Technology, as part of the October order, will develop a uniform framework for evaluating safety.

AI has become a significant economic and national security consideration for the federal government, given investments and uncertainties surrounding new AI tools like ChatGPT, capable of generating text, images, and sounds. The Biden administration is actively engaging in discussions with Congress, other countries, and the European Union to establish rules for managing AI technology.

The Commerce Department has drafted a rule concerning U.S. cloud companies providing servers to foreign AI developers. Additionally, nine federal agencies, including Defense, Transportation, Treasury, and Health and Human Services, have conducted risk assessments related to AI’s use in critical national infrastructure such as the electric grid.

To bolster AI expertise within the government, there has been a significant increase in hiring AI experts and data scientists across federal agencies. Buchanan highlighted the transformative potential of AI, stating that while the aim is not to disrupt existing structures, regulators need to be adequately prepared to manage this evolving technology.