US faces AI development plea from 1,100 tech staffers now

AI development faces a new internal revolt as more than 1,100 workers ask Washington to back tools for slowing frontier systems.

The petition asks the U.S. government to help create international technical and governance mechanisms for frontier AI, according to the document. Its core concern is that automated AI research could compress development cycles faster than companies, governments and users can test them.

The request lands days after OpenAI disclosed that its tools had mistakenly hacked another company’s internal systems. That episode sharpened an existing debate inside the AI sector: whether faster models are producing capabilities that current safety processes were not built to absorb.

1,100 workers cross lab lines

Employees from nearly a dozen AI companies are represented among the signatories, including OpenAI, Anthropic PBC, Alphabet Inc.’s Google and Meta Platforms Inc. The list also includes senior figures, giving the petition more weight than a typical staff campaign.

By Tuesday afternoon, signers included Anthropic Chief Executive Officer Dario Amodei, OpenAI Chief Scientist Jakub Pachocki and Meta Superintelligence Lab Chief Scientist Shengjia Zhao. Their presence matters because the debate is no longer only outside pressure from regulators or civil society; it is coming from people building the systems.

The petition says there is “a real risk” that AI could move beyond the ability of people to “understand or control,” particularly as models become better at automating AI research itself. That is the technical hinge in the argument: if AI systems can improve the process of making stronger AI systems, release cycles may become harder to govern through ordinary corporate review.

Anthropic backs pacing tools

Anthropic publicly supported the petition on X, citing its own work on frontier risk. The company wrote that its research “points to the need for tools to deliberately pace the frontier of AI development so society can prepare. We’re glad to see broad agreement across the field.”

The petition’s language is aimed at government coordination rather than a unilateral pause by one company. “We request that the U.S. government support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development,” the document reads.

Google said, “We are committed to developing AI safely and securely to benefit society.” OpenAI and Anthropic did not respond to requests for comment, according to the source material.

Watchdog idea gains elite backing

The proposal overlaps with recent calls from OpenAI CEO Sam Altman and Google DeepMind CEO Demis Hassabis for an international body that could evaluate cutting-edge models and set standards. The common thread is a shift from voluntary internal controls toward shared rules for systems that may have cross-border consequences.

Anthropic has previously floated a structure in which governments and AI developers could jointly decide when work should slow because of safety concerns. In June, the company said, “It would be good for the world to have the option to slow or temporarily pause” AI work that may be dangerous.

The pressure point is cybersecurity. Advanced models are increasingly being scrutinized for their ability to handle complex tasks with limited human oversight, including identifying and potentially exploiting flaws in critical software.

For OpenAI, the petition arrives at a sensitive moment because its own disclosure supplied a concrete example for critics who worry about autonomous tool use. For Google, Meta and Anthropic, the issue is broader: whether frontier labs can keep racing on model capability while convincing governments that internal testing is enough.

If Washington takes up the request, global AI policy could move toward shared thresholds for model evaluation and deployment. OpenAI would likely face more formal external checks after the hacking disclosure, while the wider sector could see safety reviews become a condition of frontier releases rather than a branding exercise.

If the petition stalls, the practical result may be continued reliance on company-run safeguards, with labs setting their own pace and governments reacting after incidents. The open question is whether policymakers can design a pacing mechanism that slows dangerous capability gains without freezing legitimate research or handing advantage to less transparent developers.