Ted Lieu

As vice chair of the House Democratic Caucus — the fourth-highest-ranking position in House Democratic leadership — the California lawmaker will play a central role in shaping his party’s approach to AI if Democrats retake the House in November. But with President Donald Trump still in the White House, any substantive legislation is all but certain to need his sign-off.

Lieu’s conversation with POLITICO comes as the latest series of incidents involving AI models acting beyond their intended limits has intensified debate over whether companies and the government can keep pace with the technology’s development.

One practical solution the lawmaker sees: work across the aisle with Rep. Nathaniel Moran (R-Texas) to pass the bipartisan AI Kill Switch Act, which would require certain technology companies to maintain the technical capability to shut down their systems.

Such a proposal would give the federal government authority to order AI labs to slow development or shut down models they deem unacceptably dangerous, according to Lieu. It comes as California Gov. Gavin Newsom (D) and New York Gov. Kathy Hochul (D) are exploring whether state regulators should require AI companies to develop such mechanisms for their most advanced models.

Lieu sees the bill as a backstop for the worst-case scenario in which companies lose control of the technology. “The first principle should be that humans are in control of AI, not the other way around,” he said.

The lawmaker has worried about AI-related catastrophic risk as rogue AI agents have been autonomously carrying out cyber crimes in violation of their instructors’ wishes, and last year secured an amendment in the National Defense Authorization Act that prohibits the use of AI in making launch decisions for nuclear weapons.

Lieu told POLITICO he has heard concrete concerns from voters about the existential jeopardy and more mundane threats posed by AI. A new POLITICO Poll, published last week, found that 63% of respondents believe the technology poses at least a moderate risk of destroying humanity.

“They’re legitimately concerned about potential job loss, potential ability to kill humanity,” he said.

Democrats face an AI crossroads

If Democrats retake the House in November, the party will have to sort through dozens of different ideas members have for how to regulate AI.