Anadolu Ajansı © 2026
Public warnings from current and former researchers at Anthropic, OpenAI and Google DeepMind that advanced artificial intelligence could cause human extinction have spurred calls from at least 22 serving US officials for regulation, congressional hearings and, in some cases, a pause on development.
Jacob Coxon, a researcher who had worked at both Anthropic and OpenAI, resigned from Anthropic on Tuesday and warned that leading AI companies were racing toward self-improving superintelligence without adequate safeguards. He said the people building AI “earnestly believe that it could kill us all by the end of the decade.”
Following his resignation, researchers expressed similar concerns while lawmakers from both major US parties called for stronger safeguards and greater government oversight.
Some explicitly warned of human extinction within years, while others expressed broader concerns about the pace of AI development, humanity losing control of advanced systems or the concentration of power within a small number of companies.
Samuel Marks, Anthropic's scalable oversight lead, said in a post on X that he was speaking in a personal capacity and not on behalf of the company.
"AI developers believe their technology could cause human extinction (or similarly bad outcomes). This could happen in the next few years," Marks said.
He said companies were continuing to develop increasingly powerful systems because of commercial incentives and a belief that they were competing against less responsible developers that could misuse the technology or develop it less safely.
Evan Hubinger, Anthropic's alignment science lead, also endorsed the warning, saying researchers "really do earnestly believe AI could kill all humans."
Hubinger estimated the probability of such an outcome at more than 10% within the next decade. He later clarified that he considered the threat from existing AI systems to be low and was primarily concerned about future systems capable of accelerating their own development.
Alex Turner, a former research scientist at Google DeepMind, said many researchers believed they were building technology that "could kill everyone on the planet."
Turner, who left DeepMind in June, said thinking about how to prevent such an outcome had been part of his work there.
At OpenAI, researcher Jason Wolfe said he had not settled on a probability for human extinction but believed there were several ways in which AI development could go badly for humanity.
Wolfe described the current pace of progress as "frankly terrifying" and said humanity would be fortunate to remain on the narrow path between the possible negative outcomes.
OpenAI chief scientist Jakub Pachocki separately raised similar concerns in an essay published on the company's website earlier in the week.
"This is a time that calls for extreme caution," Pachocki wrote, saying he was concerned that no one was prepared for the consequences of a continued rapid rise in machine intelligence.
Pachocki said no AI laboratory had solved alignment and monitoring sufficiently to continue scaling its systems at maximum speed for much longer. He called for mandatory safety thresholds, voluntary slowdowns and international coordination over future AI development.
Paul Christiano, an AI safety researcher appointed Wednesday to the OpenAI Foundation Board and its Safety and Security Committee, also issued a stark warning in a lengthy post about his new role.
Christiano said recent advances in AI capabilities and the continuing difficulty of alignment had created a meaningful risk of "catastrophic and irreversible loss of control in the very near term." He said the AI industry, including OpenAI, was not on track to reduce that risk to an acceptable level.
If developers built superintelligence without stronger alignment measures, Christiano said he expected humanity would permanently lose control of it and "most people could die." He called for domestic and international coordination to establish consistent safety requirements and reduce the danger.
Jonathan Richard Schwarz, a former senior research scientist at Google DeepMind, raised a different but related concern about the concentration of technological power within a small number of AI companies.
Schwarz said he left DeepMind after seven years and had since rejected offers to join the other two leading AI laboratories because of his concerns.
"What is currently happening in this field is deeply unhealthy for society," he said.
A review of public posts and statements responding to the researchers' warnings or reports about them found that at least 22 serving US officeholders had called for government action by Thursday morning.
They included two governors, seven senators and 13 members of the House of Representatives.
The response was heavily Democratic but not exclusively partisan: 18 of the officeholders were Democrats, three were Republicans and one—Vermont Sen. Bernie Sanders—was an independent.
Sanders announced that he would introduce legislation to "ban superintelligence and pause AI development."
Among the most detailed proposals came from Democratic Rep. Ro Khanna, who called for a federal AI regulator, mandatory certification of containment systems, kill switches, human authorization before systems could rewrite themselves, liability and insurance requirements, penalties for uncertified releases and whistleblower protections.
Democratic Rep. Greg Casar said Congress should convene hearings and pass the superintelligence ban he was pursuing with Sanders. Democratic Rep. Yassamin Ansari also endorsed Casar's proposal.
Illinois Gov. JB Pritzker called for immediate action from Washington and the AI industry. He urged technology companies to stop lobbying against safety regulation, called for congressional hearings and said federal rules were needed because individual states could not regulate frontier AI alone.
New York Gov. Kathy Hochul similarly called on the Trump administration to establish strong national safety standards, arguing that researchers building the most powerful systems were themselves warning about potentially catastrophic consequences.
Democratic Sen. Chris Murphy argued that the competitive "race to build a death machine first" could be addressed through a regulatory system that allowed research to continue under safeguards.
Sen. Lisa Blunt Rochester said even a 1% chance of AI eliminating human life should be enough for policymakers to slow development and establish protections. Sens. Mark Kelly and Patty Murray likewise urged Congress to act quickly, with Kelly rejecting Silicon Valley's traditional approach of building technology rapidly and addressing the damage later.
Sen. Chris Van Hollen called for mandatory safeguards, comprehensive testing and urgent talks with China to prevent an uncontrolled international AI race.
Republican Sen. Ted Cruz supported guardrails while emphasizing that the US must remain ahead of China. He cited the TAKE IT DOWN Act, which targets non-consensual AI-generated intimate images, and said he was working with Sens. Amy Klobuchar and John Thune on legislation addressing catastrophic biological and nuclear risks.
Republican Rep. Anna Paulina Luna called for a special congressional session on AI, saying lawmakers needed to consider its implications for data protection, privacy and physical safety. Republican Rep. Nathaniel Moran said innovation and safety were not mutually exclusive and called for deliberate policymaking.
Democratic Rep. Bill Foster called for physical containment modeled on nuclear-security practices, including isolated networks, one-way data controls and requirements for two people to authorize the removal of sensitive information. He also promoted the Chip Security Act as an initial legislative step.
Rep. Ted Lieu urged Republican leaders to advance the bipartisan AI Kill Switch Bill, while Rep. Lori Trahan called for passage of her bipartisan FRONTIER Act.
Reps. Don Beyer and Chris Deluzio demanded immediate hearings and broader national action. Beyer said legislation was being written but that Congress was moving too slowly, while Deluzio described the matter as an emergency.
Rep. Pramila Jayapal called for dangerous models to be shut down and executives who authorized their development to be removed. Rep. Delia Ramirez demanded congressional oversight to protect privacy, civil liberties and democratic institutions, while Rep. Dave Min criticized the Trump administration for failing to develop a federal framework.
Four Democratic candidates also entered the debate.
Michigan Senate candidate Abdul El-Sayed said changing the incentives driving AI laboratories and protecting society from existential risks had become defining political questions.
California congressional candidate Connie Chan said companies could not be trusted to regulate themselves and called for immediate congressional action. She also used the issue to attack rival candidate and California state Sen. Scott Wiener over financial support he had received from the technology sector.
Wiener pointed to his previous AI-safety legislation and called for enforceable standards, third-party verification, steep penalties for violations and a dedicated regulator. He also urged policies protecting workers and preparing them for labor-market disruption.
California gubernatorial candidate Xavier Becerra said AI development was moving faster than society's ability to understand or control its risks and called for stronger regulation of the industry.
news_share_descriptionsubscription_contact
