Democrats outline AI safety and liability bills for their coming congressional majority

Democrats are preparing bills on independent AI safety testing, company liability and data-center electricity costs when they take control of Congress in January 2027. The plans, described by lawmakers, aides and advocates to Reuters, put specific demands behind the incoming majority’s oversight push. President Donald Trump’s opposition to mandatory AI rules remains a major obstacle.

Three routes to greater accountability

The agenda addresses different points of contact between AI companies and the public: how powerful models are evaluated, what happens when systems cause harm, and who bears electricity costs associated with data-center growth. Lawmakers have identified priorities in each area, rather than presenting a single agreed federal AI framework.

• Independent testing: Representative Lori Trahan will pursue legislation requiring outside experts to assess leading AI models against safety standards, her deputy chief of staff, Francis Grubar, told Reuters.
• Liability for harm: Trahan also plans legislation clarifying when companies can be held responsible for harmful actions by their AI systems, Grubar said. That would address accountability after something goes wrong.
• Electricity bills: Senator Chuck Schumer said at a Washington event on November 5 that protecting consumers from higher energy costs linked to data-center expansion would be among his priorities next year.

An existing proposal shows how a liability rule might work. According to Foley & Lardner’s November 5 legal analysis, Trahan’s CLAIM Act discussion draft would create a federal right to sue developers over AI conduct that would be unlawful if a person committed it. The draft would leave state law intact. It is a proposal, not an enacted protection.

Safety proposals also reach beyond a single test. Foley describes Senator Maria Cantwell’s six-principle framework as calling for enforceable standards developed by the National Institute of Standards and Technology, independent audits before release, and mandatory incident reporting. Those elements cover both checks before deployment and disclosure of problems afterward.

responsible regulations for AI while also fostering innovation

Representative Zoe Lofgren, describing Democrats’ intended focus to Reuters

Committee power reaches beyond bill-writing

Control of congressional committees gives Democrats a platform to advance proposals and compel industry leaders to testify, even without winning agreement on new restrictions. Representative Greg Casar, chair of the Congressional Progressive Caucus, said this week that AI executives should be called to testify under oath. Democrats plan greater scrutiny of companies including OpenAI and Anthropic.

That scrutiny already has a starting point. In October, Lofgren and other Democratic House committee leaders asked OpenAI, Anthropic, Meta, Alphabet’s Google and SpaceX to explain their security practices and report instances when AI tools acted in ways their designers did not intend. The requests concern companies’ actual safety records, not just their positions on proposed regulation.

Trump and Senate rules constrain the agenda

Writing bills and passing them are different tasks. Trump has opposed binding AI mandates, and overriding a presidential veto requires two-thirds support in both chambers. Most legislation also needs 60 Senate votes to overcome a filibuster. Foley identifies disagreements among Democrats over licensing, liability and taxation as another constraint on a comprehensive package.

One possible route is attaching AI provisions to legislation Congress needs to pass, rather than seeking approval for a standalone AI bill. Raymond James policy researcher Ed Mills told Reuters that Democrats could try annual must-pass bills. Foley likewise identifies appropriations and the annual defense authorization bill as potential vehicles, while forecasting that targeted measures have better prospects than comprehensive legislation.

The incoming majority could also affect which rules Washington leaves alone. Foley forecasts that congressional efforts to override state AI laws will become less likely because Democratic proposals generally treat federal protections as a minimum, not a limit on state action. For companies, that points toward continued state compliance obligations alongside the prospect of new federal scrutiny—not a single national rulebook replacing them.