Employees from OpenAI, Anthropic, Google, Meta, Thinking Machines, Microsoft, and Mistral have signed a letter urging the US government to address the risks posed by automated AI research. The group warns that while artificial intelligence could build a better future, that result is not guaranteed. They note that leading companies might soon automate the creation of new AI systems, which could accelerate progress beyond current human oversight. The statement argues that this rapid capability development creates a real risk that requires immediate attention from regulators. The signatories are asking for a temporary slowdown in frontier model training or at least a faster pace of global coordinated governance. They believe current safety measures are insufficient against an automated research loop.
This move marks a rare moment where industry builders publicly admit they cannot fully control their own output. Previously, companies often framed safety as a competitive advantage or a solved engineering problem. Now they suggest the technology may outpace human ability to manage it, forcing a shift from voluntary guidelines to mandatory rules. The letter implies that without intervention, the speed of innovation could lead to unintended consequences that no single firm can mitigate alone.
- The signatories represent major US and European labs including Mistral.
- The letter specifically targets US federal agencies for action.
- The group proposes either a development pause or accelerated international rules.
