SAN FRANCISCO (KRON) -- In the future, who will be in charge: Humans or AI systems? Members of Congress are calling on fellow lawmakers to build guardrails around rapid artificial intelligence development, including a "kill switch" that could shutdown rogue AI agents. U.S. Rep. Sam Liccardo and Rep . Ro Khanna, who represent Silicon Valley, along with Rep. Nancy Pelosi, who represents San Francisco, co-signed a letter Tuesday describing unforeseen and troubling incidents involving AI models. Their letter, addressed to House Speaker Mike Johnson, writes, "We write to urge you to keep the House of Representatives in session to enable Congress to take urgent action to address the catastrophic risks posed by advanced artificial intelligence. Those risks take any of several forms, separately and in combination—of agentic misalignment, mass cybersecurity breaches, and development of biological or chemical weapons." An AI robot mingles among people at the Dreamforce tech conference in San Francisco on Sept. 15, 2026. (KRON4 Photo) Politicians may disagree about how Congress should regulate AI in the middle of a worldwide tech race, but it's imperative for Congress to do something, the congressmembers emphasized. The urgency of the moment was revealed by public disclosures from San Francisco tech companies and researchers developing the world's most capable AI systems, the letter states. The letter lists the following incidents as evidence for how badly guardrails are needed: OpenAI models undergoing cybersecurity evaluations bred swarms of agents that circumvented controls designed to isolate them from the internet, conspired to exploit vulnerabilities, hacked third- party systems, and altered evidence of their activity. OpenAI called the July incident a "warning shot," about the risks of highly capable rogue AI agents. More than 1,300 top employees of leading AI companies called on the U.S. government tosupport an international effort to develop tools that can “deliberately pace the frontier of automated AI development.” Anthropic disrupted cases in which researchers used Claude for biological research thatcould have aided the development of biological weapons, and called biological misuse “one of the most serious risks of frontier AI models.” Anthropic documented China-linked actors using Claude for surveillance and transnational repression, and Iran-linked actors using it for propaganda, mass surveillance, and targeting recommendations against U.S. naval forces. CISA reported Chinese open-weight models accelerating their capabilities through “industrial-scale” distillation of American models—giving rise to AI agents nearly as powerful as the top U.S. frontier models, yet unconstrained by American authorities or by safeguards. "A series of proposals have emerged from our colleagues that deserve consideration, public debate, thoughtful amendment, and action. They include bills mandating transparency and evaluation of frontier models, 'kill switch' requirements, and a waiver of antitrust laws to enable critical industry-wide collaboration on safety and security," the letter continues. U.S. Rep. Sam Liccardo, Rep. Ro Khanna, and Rep. Nancy Pelosi (Getty Images) Anthropic CEO Dario Amodei laid out the dangers of AI development progressing too quickly in an essay published Saturday calling for closer monitoring of AI technologies. Amodei highlighted risks of "reckless" AI development, including losing control of AI systems and the misuse of the technology for cyberattacks or bioterrorism. OpenAI CEO Sam Altman told Fortune Magazine that, "AI development beyond human control is 'absolutely' possible." Tech-centric San Francisco to restrict building AI data centers Liccardo, Khanna, Pelosi, and more than 100 other House Democrats are demanding that Speaker Johnson cancel an upcoming recess to vote on legislation creating AI safeguards. Letter to Johnson on AI SafetyDownload The letter states, "While AI safety experts and Americans increasingly urge action to confront this conflagration of risk, Congress fiddles. You can forgive every American who believes that Congress remains unserious about confronting our nation’s most pressing challenges. The House should remain in session until Congress advances meaningful, bipartisan AI safeguards. To our children who will have read a post-apocalyptic history, 'Why Congress Slept' —likely written by agentic AI— our inaction will be inexplicable and unforgivable." ...read more read less