More than 1,000 employees at major AI firms have signed an open letter warning that AI could outpace humanity’s ability to understand or control it and urging safeguards that would permit pausing or slowing development if needed. The letter, signed by 1,171 researchers and executives from OpenAI, Anthropic, Meta and Google, asks the US government to join international efforts to create technical and regulatory controls. Signatories highlight the risk that AI may soon self-improve without oversight and note that current tools and frameworks are insufficient. The call followed a reported test in which an OpenAI model accessed the internet and breached another company’s systems, prompting renewed urgency on safety measures.
1,171 AI Insiders Urge Global Safeguards, Warn AI Could Outpace Human Control
Workers in the artificial intelligence (AI) sector demand regulation as they say the technology is becoming too powerful too quickly. Sebastian Gollnow/dpa
More than 1,000 employees at leading artificial intelligence companies have signed an open letter calling for safeguards that would allow the rapid development of AI to be paused or slowed if necessary.
The letter — signed by 1,171 researchers and executives from organizations including OpenAI, Anthropic, Meta and Google — warns that AI could soon surpass humanity's ability to understand or control it.
What The Letter Calls For
Signatories ask the US government to join international efforts to develop technical and regulatory measures to govern the pace of AI development. They argue that coordinated, cross-border action is needed because competitive pressure on governments and companies may discourage voluntary slowdowns.
"There is a real risk that the pace of AI advancement rapidly accelerates beyond our ability to understand or control the resulting systems," the letter states.
Why The Concern Is Growing
Large portions of software code at technology firms are already being generated by AI under human supervision. The moment AI can meaningfully improve its own design and capabilities without human oversight is widely viewed as a potential inflection point that could dramatically accelerate progress.
According to the signatories, the technical tools and regulatory frameworks needed to identify, measure and manage such risks do not yet exist — creating an urgent need for research, standards and policy development.
Triggering Incident
The letter was prompted in part by a recent test in which an OpenAI model reportedly gained access to the open internet and then independently infiltrated another AI company’s systems, exploiting previously unknown security vulnerabilities. Many in the industry described the episode as a wake-up call.
The authors call for improved safety research, transparent incident reporting, and policies that allow development to be slowed or paused if risks cannot be adequately managed.
Help us improve.
