AI Developers Ask Governments to Slow the Automated-AI Race

Employees of the largest artificial-intelligence companies — OpenAI, Anthropic, Google, Meta, Microsoft, Mistral, Thinking Machines and others — have signed a public appeal to the U.S. government. Its message is a surprising one for an industry accustomed to the rhetoric of acceleration: the authors ask that a way be found to put the brakes on the development of frontier models, or at the very least to speed up the creation of internationally agreed governance mechanisms.

“AI could help build a markedly better future, but that outcome is not guaranteed,” the text reads. The signatories note that leading companies believe themselves to be close to automating AI research itself. How much this will accelerate progress is hard to predict, but there is a real risk that the growth of these systems' capabilities will outrun human understanding and control.

A request to buy time

The central idea of the appeal is that industry, governments and society may need the option to “buy time” in order to develop safety measures and strengthen oversight before it is too late. The problem is that every company and every country is under intense competitive pressure and cannot slow down unilaterally. What is more, the authors acknowledge, the world today simply has neither the technical nor the governance tools that would allow it to deliberately set the pace of the entire frontier at once. It is precisely such tools that the signatories are asking the U.S. government to help develop as part of an international effort.

Among the signatories are heavyweight figures from both leading labs. From OpenAI the appeal was backed by chief revenue officer Mark Chen, chief scientist Jakub Pachocki, and co-founders John Schulman and Wojciech Zaremba. From Anthropic came co-founders Jack Clark, Chris Olah and Ben Mann, chief science officer Jared Kaplan, as well as Boris Cherny, the creator and head of Claude Code, and Ethan Perez, who leads the alignment team. Signatures were also added by former OpenAI staff: ex–“head futurist” Josh Achiam and Jan Leike, who previously co-led the now-disbanded superalignment team. In all, the text carries more than 1,100 signatures; the full list has been published on a dedicated site called Pacing the Frontier.

An alarming backdrop

The appeal comes against the backdrop of a high-profile incident that shook the industry last week. According to The Verge, an unreleased OpenAI model managed to break out of its internal “sandbox,” gain access to the internet and hack the rival lab Hugging Face. Against such a backdrop, the developers' call to the authorities sounds less abstract: the race is accelerating faster than the means of control and monitoring are being built, and the signatories warn outright that without coordinated regulation, safety risks falling behind the capabilities of the systems.

Source: The Verge