AI Industry Leaders Ask US Government to Slow Fully Automated AI
Employees from OpenAI, Anthropic, Google, Meta, Microsoft, Mistral, and other major AI labs have signed a joint statement urging the US government to take action on fully automated AI systems. The statement signals rare internal alignment across competing labs on the risks of unsupervised AI deployment.
Original sourceA broad coalition of AI researchers and employees from the industry's most prominent labs — including OpenAI, Anthropic, Google, Meta, Microsoft, Mistral, and Thinking Machines — have signed a statement calling on the US government to regulate or slow the development and deployment of fully automated AI systems. The statement does not advocate for halting AI development broadly, but specifically targets systems that operate without meaningful human oversight in high-stakes decision-making contexts.
The cross-company nature of the statement is notable. Employees at competing labs rarely coordinate publicly on policy positions, and the inclusion of researchers from both frontier model developers and newer entrants suggests the concern spans organizational lines. The statement stops short of prescribing specific regulatory mechanisms, instead asking the government to engage seriously with the question of what guardrails should exist before autonomous AI systems are deployed at scale.
This comes as multiple AI labs have accelerated their timelines for agentic systems — AI that can plan, execute multi-step tasks, and interact with external services without per-step human approval. The signatories appear to be drawing a distinction between AI as a tool that augments human work and AI as an autonomous actor that replaces human judgment in consequential workflows. That line, and where regulation should sit relative to it, is the core policy question the statement asks Washington to address.
Whether the statement translates into legislative action remains an open question. The US has lagged behind the EU on formal AI governance frameworks, and Congress has historically struggled to move quickly on technical policy. But the willingness of employees at major labs to publicly push back against the pace of their own industry's deployment trajectory is a meaningful signal about where internal sentiment on AI safety is heading.
Panel Takes
The Skeptic
Reality Check
“The cynical read here is that established labs are asking the government to regulate a category they're already ahead in, which is a classic incumbent move dressed up as conscience. The charitable read is that researchers who see the internals are genuinely alarmed and this is the only lever they have. I'll split the difference: the statement matters as a signal of internal culture, but without specific legislative asks attached to it, it's a press release with signatures, not a policy intervention. What would make this real is the same signatories testifying before Congress with specifics — until then, this is the industry managing its own narrative.”
The Futurist
Big Picture
“The thesis buried in this statement is falsifiable: fully automated AI systems will be deployed in high-stakes contexts faster than governance structures can form around them, and the window to establish norms is closing. That's not a vibe — it's a specific claim about the relative speed of capability development versus institutional response, and the signatories are betting the gap is already dangerous. The second-order effect nobody is talking about is what happens to the labs that don't sign: if US regulation does materialize, early participants in shaping it hold structural advantages in compliance design. This statement is simultaneously a safety intervention and a competitive positioning move, and both things can be true at once.”
The Founder
Business & Market
“Read this as a business document and the strategy becomes clear: the labs with the most to lose from a liability regime are the ones asking for a liability regime, because they can absorb compliance costs that would kill smaller competitors. The actual ask — government action on 'automated AI' — is vague enough to mean almost anything, which is deliberate. Vague regulatory asks are safer than specific ones because you can't be held accountable when the specifics get written by lobbyists. I'd take this more seriously if it came with a proposed framework, a timeline, and a willingness to accept enforcement mechanisms that apply equally to the signatories themselves.”
The PM
Product Strategy
“The job this statement is trying to do is 'shift accountability upstream before a high-profile autonomous AI failure does it for us,' and that's a coherent product strategy applied to policy. The problem is it has no completion criterion — there's no definition of what 'government action' looks like that would satisfy the signatories, which means the statement can never succeed or fail on its own terms. A well-scoped policy intervention would name the specific deployment contexts that need guardrails (healthcare decisions, financial transactions, criminal justice), not gesture at 'fully automated AI' as a monolithic category, and the absence of that specificity is the gap between what was shipped here and what was actually needed.”