US Senators Mark Warner and Thom Tillis have put forward a proposal aimed at monitoring security issues by creating a repository to document breaches in artificial intelligence systems.
"Significant vulnerabilities in AI algorithms can differ significantly from vulnerabilities in traditional software, requiring a rethinking of the government's approach to addressing such issues," the lawmakers said.
The proposal would establish an Artificial Intelligence Security Center at the U.S. National Security Agency to serve as a hub for AI research for both the private sector and academia, much like a testing ground.
It also proposes to develop protocols to prevent or mitigate strategies to counter AI and ensure the safe deployment of this technology.
The proposal directs the U.S. National Institute of Standards and Technology (NIST) to revise the National Vulnerability Database (NVD), a government database containing information on software and hardware vulnerabilities.
"The goal is to improve the monitoring and management of AI-related security incidents,"the proposal emphasizes.
According to the draft, NIST is tasked with updating the NVD to account for differences between AI systems and traditional software, including differences between common exploits and those designed to subvert algorithms.
In addition, the U.S. Cybersecurity and Infrastructure Protection Agency will need to modify the existing database of known vulnerabilities (CVEs) or create a new mechanism to identify AI security weaknesses. Each vulnerability is assigned a CVE number that is integrated into the NVD.
The #artificial intelligence Security Act must go through committee vetting before reaching the U.S. Senate for consideration.
Earlier, US Senators Kirsten Gillibrand and Cynthia Lummis introduced a bill to regulate stablecoins.
