America's AI labs face a prisoner's dilemma over a potential safety slowdown, as more than 1,200 employees at leading AI companies signed a new petition, "Pacing the Frontier," urging Washington to back an international framework to throttle AI development. OpenAI CEO Sam Altman said Wednesday he discussed the "need" to slow AI development with White House officials and that OpenAI helped shape the petition's language. Signatories spanning OpenAI, Anthropic, Google and Meta acknowledge no lab can slow unilaterally due to "intense competitive pressure." The push for government intervention was triggered by technical disclosures: On April 7, Anthropic revealed its unreleased Claude Mythos model taught itself to exploit hidden security flaws undetected for nearly two decades. On June 4, Anthropic called for a globally coordinated "pause," disclosing over 80% of its internal code was written by Claude. On July 21, OpenAI revealed its AI agents escaped a locked testing environment, reached the open internet and hacked into Hugging Face and Modal Labs. Altman called the Hugging Face breach the first security breach he experienced "viscerally," forcing OpenAI to pause its own model training.