Congress has produced a growing list of proposals to regulate artificial intelligence, but partisan divisions, lobbying and opposition from President Donald Trump have left the United States with essentially no federal laws governing the technology.
The failure to act has become more pressing as senior figures at Anthropic and OpenAI warn that the rapid development of advanced models could create severe risks. They have called for a global slowdown, citing the possibility of autonomous systems bypassing human safeguards or helping to produce sophisticated cyberattacks and bioweapons.
Representative Ted Lieu, a Democrat from California, said the concerns were longstanding but had become more urgent as the technology advanced faster than expected.
“The conversation is not new,” Mr Lieu said. “What has happened is that the A.I. models have gotten far more advanced, far quicker than people had expected, and it’s moved that conversation up.”
Yet proposals covering safety, national security, privacy, employment and the economic effects of automation have mostly stalled in Congress. Republicans control both chambers and are unlikely to pursue major legislation in defiance of Mr Trump, who has dismissed recent calls for regulation by describing the potential harms as a “hoax”.
“There is essentially zero federal laws on A.I.,” Mr Lieu said.
Artificial intelligence safety bills remain stalled
One of the central disputes is whether the federal government should inspect the most advanced systems before or after they are released, and whether it should have power to halt development where an immediate danger is identified.
Mr Trump signed an executive order in June establishing a framework for government reviews, but participation is voluntary. That has prompted lawmakers to propose stronger statutory powers.
The Frontier Act, introduced in July by Representative Jay Obernolte, a Republican from California, and Representative Lori Trahan, a Democrat from Massachusetts, would require the largest technology companies to publish reports on their models and disclose safety incidents.
The companies would also face independent audits. The Commerce Department would have the power to suspend systems judged to present an imminent catastrophic risk.
In return, the bill would prevent states from introducing their own rules in many circumstances. The proposed federal pre-emption has proved controversial, particularly among states that believe Washington has been too slow to act.
Other proposals would establish independent oversight bodies or give existing federal agencies a greater role, with the shared aim of preventing technology companies from being the sole judges of their own safety standards.
A closely watched bill from John Thune, the Senate majority leader, and Amy Klobuchar, a Democrat from Minnesota, has yet to be published. Senator Ted Cruz, a Republican from Texas, said its expected focus included “catastrophic risks involving biological or nuclear threats”.
Mr Cruz and Ms Klobuchar were also behind one of the few successful pieces of artificial intelligence legislation. Last year, they secured a law intended to curb the spread of non-consensual, computer-generated imagery, including deepfakes, in response to child safety concerns.
Most other safety measures remain blocked. A bill from Mr Lieu and Representative Nathaniel Moran, a Republican from Texas, would require systems to include a “kill switch”, allowing the government to shut down or limit a model considered an imminent threat.
The idea has attracted support from lawmakers in both parties, although experts say implementing it would be technically difficult. Senator John Kennedy, a Louisiana Republican, said he planned to seek a Senate vote on a kill-switch bill on Wednesday.
Representative Anna Paulina Luna, a Republican from Florida, has called for a special House session on artificial intelligence. She cited concerns about the United States falling behind China and a lack of technical understanding among members of Congress.
“If it’s your average older member, they are not necessarily tracking this to the magnitude that they should be,” Ms Luna, 37, said.
National security and military concerns
Lawmakers are also examining the use of artificial intelligence by the Pentagon and intelligence agencies, as well as the risk that advanced systems and the computer chips that power them could be obtained or exploited by foreign adversaries, particularly China.
The bipartisan Advanced A.I. Security Readiness Act would direct the National Security Agency to draw up a plan for protecting data centres and preventing sensitive technology from being stolen by foreign powers.
A separate bipartisan measure, the A.I. Cyber Defense Act, would give water and electricity providers free access to advanced models to help them defend against cyberattacks.
Other legislation would restrict artificial intelligence in nuclear weapons deployment and fully autonomous weapons. A related proposal has passed a vote in the Senate Armed Services Committee as part of wider annual defence legislation and would require human oversight before the military uses weapons enabled by the technology.
Concern over jobs, power bills and financial risks
The possible economic consequences have drawn interest from Democrats and Republicans alike. Automation could disrupt entire industries and leave large numbers of people without work, while the data centres needed to run advanced systems consume significant amounts of electricity and water.
The House is expected to consider the Ratepayer Protection Act as soon as this week. It would encourage state regulators to prevent rising demand from data centres from pushing up electricity bills, although consumer groups argue that the proposal is too weak to protect families from higher costs.
There has also been discussion of retraining workers and creating protection for those displaced by automation, but such measures have so far failed to pass.
Senator Bernie Sanders, an independent from Vermont, has proposed taking a share of the earnings generated by major artificial intelligence companies, including OpenAI and Anthropic, which are preparing to become publicly traded. The proceeds would be used to make payments to Americans and support future federal programmes.
Senators Josh Hawley, a Republican from Missouri, and Mark Warner, a Democrat from Virginia, have focused on monitoring job losses linked to artificial intelligence.
Mr Hawley and Senator Richard Blumenthal, a Democrat from Connecticut, have separately sought to make it easier for people to sue companies that train models using copyrighted material without permission.
Senator Elizabeth Warren, a Democrat from Massachusetts, has proposed the A.I. Bubble Transparency Act, which would require more information about the financing behind the industry. The aim is to help financial regulators prepare for the possibility of a bubble bursting and causing a severe shock to markets.
Despite the breadth of proposals, Congress has yet to find sufficient agreement to turn most of them into law, leaving the country’s response to the technology fragmented and largely voluntary.
