Autonomous cyberattacks fuel US Congress' AI focus through 2026 midterms and beyond

Roughly two months before the U.S. midterm elections, the contours of artificial intelligence policy debates at the federal and state levels are becoming more clear. Lawmakers on both sides of the political spectrum are floating policy proposals amid growing concerns around developers losing control of frontier models following recent high-profile cyber incidents. 

Speaking exclusively to Semafor 10 Sept., sources with knowledge of negotiations between U.S. Sens. Amy Klobuchar, D-Minn., Ted Cruz, R-Texas, and John Thune, R-S.D., said they are working on a bipartisan AI safety bill that was motivated largely by the resignation of Anthropic researcher Jacob Coxon, who warned AI labs' race to develop self-improving models may usher in the extinction of the human race. The sources said they believe the bill represents the only realistic shot at passing AI safety legislation prior to the midterms and could be introduced as early as next week. 

Klobuchar said in a statement to Semafor the legislation would offer "commonsense guardrails to address the risks posed by AI systems."

The behind-the-scenes negotiations among Klobuchar, Cruz and Thune come as OpenAI is both publicly and privately lobbying Congress to act on establishing a national AI safety framework. 

On 9 Sept., OpenAI Chief Global Affairs Officer Chris Lehane wrote on the company's website about the outreach efforts to establish both national and international AI safety standards. Lehane highlighted four components of OpenAI's strategy: Lobbying Congress for mandatory national AI safety requirements, supporting "state legislation that strengthens the broader AI safety ecosystem," collaborating with other frontier AI labs to establish "industry-led standards" and building global standards "even if that means slowing the advancement of model capabilities."

"We've reached a new chapter in AI capabilities, and that demands a new chapter for AI policy," Lehane wrote "No company, industry, or government can meet this challenge alone. We need to meet this moment with a bias toward meaningful action over policy perfection."

Additionally, Punchbowl News reports OpenAI contacted a number of congressional offices 9 Sept. urging members to "act now to pass foundational AI safety legislation," while pledging to "work with Congress to meet the moment," according to an email reviewed by the publication. 

California AI safety bills covering audits, independent safety assessments

At the state level, California is positioning itself to lead on AI rules in the same way it has done previously with privacy legislation.

Gov. Gavin Newsom, D-Calif., recently enacted a pair of landmark AI safety bills, Senate Bill 813 and Assembly Bill 1405, following passage by the state legislature. 

SB 813 requires the state Government Operations Agency to designate "independent verification organizations" that will assess potential risks posed by new frontier models released on the market by 1 Jan. 2028. AB 1405 requires the GOA to establish an AI Auditor Registry by 1 Jan. 2029 that is designed to protection against unregistered entities from "offering, selling, or conducting a covered AI audit."

However, neither of the bills signed by Newsom contain a so-called "kill switch" requirements on developers, which were contained in the previously passed SB 1047 that was vetoed two years ago. U.S. Rep. Ted Lieu, D-Calif., reiterated the need for kill switch provisions in federal AI safety legislation in wake of this week's resignation of the Anthropic researcher. 

Advocacy groups seek transparency from White House on voluntary AI safety framework

With the potential for state legislatures to further address AI development and use in a fragmented manner, as the drumbeat for Congress to step in is growing louder.

More than two dozen advocates for AI safety on both the left and right ends of the political spectrum are calling for the Trump administration to release information related to its voluntary AI security framework, according to Semafor. Last month, the White House shared high-level details of the framework with a handful of technology companies, but specifics have yet to be released publicly.

"The government has a fundamental responsibility to share the parameters of the process for reviewing frontier AI models prior to wider deployment," according to the stakeholder letter, led by Americans for Responsible Innovation and the Center for Democracy and Technology. 

Additionally, Semafor reports nonprofit Protect Democracy filed a lawsuit against the Trump administration seeking a federal court to enforce a Freedom of Information Act request to obtain the framework. 

What is top of mind for Congress?

While the possibility of passing some form of AI safety legislation prior to the November election may be remote, federal lawmakers from both sides of the aisle are keeping AI near the top of the agenda throughout the fall. 

In a letter obtained by Axios, Sen. Josh Hawley, R-Mo., wrote to OpenAI CEO Sam Altman regarding plans for the Senate Committee on Homeland Security and Governmental Affairs Subcommittee on Disaster Management to open an investigation into OpenAI's role in the Hugging Face cyberattack in July. Hawley's letter seeks answers to 16 questions related to the attack and how OpenAI responded by 1 Oct. 

"The American people deserve to know the details of what went on in the Hugging Face incident and other incidents of AI models going rogue," Hawley wrote, describing OpenAI's test that precipitated the attack and the company's reported failure to implement mitigation measures upon discovering the incident as "reckless." 

Meanwhile, Sen. Bernie Sanders, I-Vt., is reportedly planning to introduce his own bill that would ban AI labs from developing super intelligence and pause AI development more broadly, according to Semafor. 

Additionally, Sanders will hold a bipartisan briefing with fellow lawmakers next week that is anticipated to address "extraordinary dangers" of AI, Axios reports. The closed-door session will reportedly feature University of Toronto Professor Emeritus in Computer Science Geoffrey Hinton, Future of Life Institute co-founder Max Tegmark and independent researcher Ajeya Cotra who took part in an investigation into the Hugging Face hack. 

"Congress needs serious discussions about this issue," Sanders wrote in a letter to colleagues. 

Looking beyond the midterms, Politico reports House Democrats are beginning to prepare for establishing a select committee on AI should they regain control of the chamber. The committee would have subpoena power and House Minority Leader Hakeem Jeffries, D-N.Y., is reportedly open to creating it. 

The push for an AI select committee is led by Rep. Bill Foster, D-Ill., as well as Rep. Lieu. However, one factor that could prove difficult to get the committee off the ground is garnering support among Democrats who sit on committees with existing AI oversight roles. 

"I am supportive of a select committee on AI," Foster told Politico in a statement. "When I’m back home talking to voters, AI is consistently one of the top issues — if not the top issue — on their minds. Leader Jeffries understands that we can't repeat Republicans' mistake of letting Congress sit on the sidelines while AI rapidly shapes our future."