Over 1,100 workers at OpenAI, Google, Meta, Anthropic, and other AI companies have signed an open letter to the US government demanding federal regulation of their own employers’ artificial intelligence systems.
The petition represents an extraordinary moment of internal dissent in an industry racing to build ever-larger AI models with minimal external oversight. These are not external critics or academic researchers—they are the engineers, researchers, and product managers actually building the systems they want regulated. They work inside the companies with the most at stake in any regulatory framework.
- Scale of Dissent: More than 1,100 employees across OpenAI, Google, Meta, Anthropic, xAI, and Stability AI have publicly signed a letter demanding federal regulation of their own employers.
- What They Are Demanding: The letter calls for mandatory pre-release safety testing, training data disclosure requirements, independent audits in high-stakes domains, and formal whistleblower protections for AI workers.
- The Structural Argument: Signatories argue that without binding rules applying equally to all companies, competitive pressure to ship fast will always override the incentive to test thoroughly.
The signatories, whose names and affiliations were published as part of the letter, come from companies including OpenAI, Google, Meta, Anthropic, xAI, Stability AI, and dozens of smaller AI labs. The letter calls on Congress and federal agencies to establish binding safety standards, transparency requirements, and accountability mechanisms for advanced AI systems before they cause irreversible harm.
“We are employees at leading AI companies and other signatories who are deeply concerned about the risks posed by AI systems currently being developed and deployed,” the letter states. The workers are asking for regulation they describe as necessary to protect the public from potential harms.
Why Are the People Building AI Asking to Be Regulated?
This petition breaks a pattern that has defined the AI industry since its recent acceleration: company leadership publicly resists regulation while privately acknowledging risks, and employees stay silent. The signers are betting their professional reputations that speaking out is more important than loyalty to their employers’ preferred policy positions.
The letter specifically calls for mandatory safety testing before AI systems are released to the public, requirements that companies disclose their training data sources and methods, and independent audits of AI systems used in high-stakes domains like criminal justice, hiring, and healthcare. It also demands that workers at AI companies be protected if they report safety concerns internally or externally—a direct reference to the lack of whistleblower protections in the sector.
The pattern here is not without precedent in the technology industry. The Cambridge Analytica scandal demonstrated what happens when powerful data systems operate without external accountability structures: internal employees who understood the risks stayed quiet, self-regulatory assurances proved hollow, and the public absorbed the consequences. The 87 million Facebook users whose data was harvested for psychographic profiling had no meaningful recourse because no binding framework required disclosure, audit, or consent. The AI workers signing this letter appear to have absorbed that lesson directly—they are not waiting for the equivalent crisis to force the conversation.
Anthropic, one of the signatory companies, has publicly stated it supports AI regulation. But even at Anthropic, employees felt compelled to add their names to a letter demanding their own company comply with stronger rules. OpenAI and Google have taken more cautious public stances on regulation, arguing that industry self-governance should come first. The gap between those positions and what their own workforces are demanding is now a matter of public record. For a deeper look at how AI training data practices already operate with minimal oversight, the pattern of opacity these workers are describing is already visible in current deployments.
• 1,100+ signatories spanning OpenAI, Google, Meta, Anthropic, xAI, Stability AI, and dozens of smaller labs
• No comprehensive federal AI regulation has passed in the United States despite multiple congressional hearings over two years
• AI systems are currently deployed in medical diagnosis, loan decisions, and content moderation with no mandatory independent audit requirement
• California has passed narrower state-level bills targeting deepfakes and biased hiring algorithms, but a national framework remains absent
What Do the Workers Actually Know That the Public Does Not?
What makes this petition distinct is that it comes from people with direct knowledge of how these systems actually work. A software engineer at Google who helped build a large language model knows the gaps between public claims about safety and what happens in practice. A researcher at OpenAI understands the difference between a system tested in a lab and one rolled out to millions of users. These workers are not speculating about risks—many have seen them firsthand.
One dimension of that insider knowledge concerns reliability in high-stakes contexts. Research from Stanford’s Human-Centered Artificial Intelligence institute has documented disturbing and pervasive errors among leading large language models across a range of legal tasks—errors that would be consequential if the systems producing them were deployed in criminal justice or regulatory contexts without independent review. The petition’s demand for audits in exactly these domains reflects awareness of this gap between laboratory performance and real-world reliability.
The algorithmic transparency problem compounds this. Even when errors occur, the opacity of how these systems reach their outputs makes accountability nearly impossible without formal audit mechanisms. The workers signing this letter are not simply asking for rules—they are asking for the infrastructure that would make those rules enforceable.
• Stanford Law School’s Regulation, Evaluation, and Governance Lab has assessed leading AI legal research tools and found significant reliability concerns, raising questions about deployment in consequential decision-making contexts without mandatory testing standards
• The absence of pre-release safety testing requirements means companies currently self-certify the readiness of systems before public deployment
• Independent audits of the kind the petition demands do not currently exist as a legal requirement for any AI system deployed in the United States
Is the Competitive Race to Deploy AI Making Everyone Less Safe?
The letter also addresses a structural problem in AI development: companies compete on speed and capability, not safety. If one lab slows down to conduct more rigorous testing, another lab races ahead. Regulation that applies equally to all companies would level that playing field and remove the incentive to cut corners. The signatories appear to believe that their own companies’ long-term interests align with this kind of regulation, even if short-term growth might slow.
The petition does not name specific incidents or leaked documents. Instead, it frames regulation as inevitable and argues that shaping it now is better than facing it later under crisis conditions. The letter warns that without proactive rules, a major AI failure could trigger a regulatory backlash that is far more restrictive than what careful policymaking would produce. This argument—that industry actors should prefer structured regulation over crisis-driven restriction—mirrors the logic that eventually drove parts of the financial sector to accept post-2008 oversight frameworks.
For users interacting daily with AI personal assistants through ChatGPT, Google’s Gemini, or Meta’s AI tools, this petition signals something significant: the people building these tools believe they need external constraints. It is an admission that internal safeguards alone are insufficient. The systems shaping search results, drafting legal summaries, and informing medical decisions are built by people who now say those decisions should be subject to government oversight.
What Happens to the Workers Who Signed?
The petition also highlights a generational divide in tech. Older tech workers remember the early internet era when companies self-regulated and faced minimal government scrutiny. Many of the AI workers signing this letter appear to believe that era was a mistake—that waiting until a technology causes widespread harm before regulating it is reckless.
How will company leadership respond? Some, like Anthropic, may quietly welcome the letter as evidence that their regulatory position is reasonable. Others, particularly OpenAI and Google, face pressure to either endorse the petition or explain why their employees are wrong to demand it. Publicly rejecting your own workforce’s safety concerns is a difficult position to sustain, particularly when those concerns are now documented and attributed.
• The petition’s demand for whistleblower protections reflects a documented gap: AI workers currently have no formal legal shield if they report safety concerns externally, meaning internal dissent is the only available channel
• The framing of regulation as a competitive leveling mechanism—rather than a burden—represents a strategic argument designed to neutralize industry opposition from smaller labs that fear compliance costs
• Congressional use of insider testimony has historically accelerated regulatory timelines; the public attribution of signatories gives lawmakers a politically durable citation that is difficult for industry lobbyists to dismiss
Will Congress Act Before the Crisis Forces Its Hand?
Congress will likely cite this petition in future AI regulation debates. Lawmakers can now point to insiders at the world’s most powerful AI companies and say: even the people building these systems want you to regulate them. That is a politically powerful argument that is hard for industry opponents to dismiss, particularly when the signatories are named and their affiliations are verified.
The real test comes next. Will the signatories face retaliation for speaking out? Will their companies actually support the regulations they are demanding? And will Congress act on this rare moment of internal industry agreement that something needs to change? The petition is public. The pressure is now visible. What happens next depends on whether these 1,100 workers have genuinely shifted the conversation, or whether they have simply made their dissent official before being ignored.
