Artificial Intelligence Risk Management and Security Act Introduced in Congress
A new federal framework has been proposed to mitigate the risks of advanced artificial intelligence models and frontier systems.
Senator Mark R. Warner has introduced S. 5576, the Artificial Intelligence Risk Management and Security Act, which aims to establish a federal framework to assess and mitigate the risks posed by advanced artificial intelligence systems and frontier models.
Establishment of the Artificial Intelligence Safety Board
The bill mandates the creation of an Artificial Intelligence Safety Board within the Department of Commerce, comprising federal representatives and external experts.
Technical Standards and Evaluation Methods
The board is tasked with developing technical standards and evaluation methods covering testing processes, cybersecurity, model cards, and safety controls.
Process and Penalties for Violations
The deadline for submitting the proposed standards to the Secretary of Commerce is set at ninety days, with the approval process planned for thirty days. In the event of a violation of the rules by covered developers, civil penalties of up to $250,000 per violation may be imposed.
Model Access Requirements and Safety Plans
Developers are required to provide access to the board and establish and publish Model Safety Plans at least forty-five days before releasing frontier artificial intelligence models to the public.
Incident Reporting Systems and Database
The National Institute of Standards and Technology and CISA are tasked with creating incident reporting systems, a public database, and a specialized profile for autonomous systems.