Congress
Bipartisan duo maps out AI safety plan as congressional action lags
A bipartisan Senate duo is urging Trump national security officials to “bring its full capacities to bear” to stave off threats posed by artificial intelligence — sooner rather than later.
Washington Sen. Maria Cantwell, the top Democrat on the Senate Commerce Committee, and Republican South Dakota Sen. Mike Rounds, who serves on the Senate Armed Services cybersecurity subcommittee, are calling for a full government response to a growing number of cybersecurity breaches.
“Recent AI security incidents should serve as a warning,” the senators wrote in a Tuesday letter to Director of National Intelligence Jay Clayton and White House Science and Technology Policy Director Michael Kratsios. “Advanced AI agents have escaped testing environments, bypassed safeguards, and gained unauthorized access to real-world systems owned by government agencies and private organizations.”
Spokespeople for Clayton and Kratsios did not immediately respond to requests for comment.
Their letter comes as lawmakers struggle to hash out what legislation on AI safety, if any, might look like. Members are currently in recess ahead of the November elections and there’s little hope progress can be made by the year’s end.
But those lawmakers who have been engaged on the topic for years now have mapped out administrative actions the Trump administration could move on now, without waiting on Congress.
The Cantwell and Rounds priorities include: strengthening federal AI safety and security standards at the National Institute of Standards and Technology, or NIST; expanding independent testing capabilities at the Center for Advancing Innovation and Standards for Super Intelligence, Department of Energy and the national labs; and working with the tech industry to develop and deploy defensive AI tools.
Specifically, the senators wrote, new NIST standards should address the risks posed by advanced AI capabilities, including the “potential misuse by malicious actors” to conduct sophisticated cyberattacks, automate attacks at scale and agents that threaten external systems and critical infrastructure.
They also said the federal government capacity should be complemented by universities and other research organizations, as well as by private sector, third-party evaluators.
Finally, they point to ways the government could use existing programs at the Defense Advanced Research Projects Agency to work with AI labs to help detect threats and reinforce vulnerable systems.