On July 28, 1,178 AI professionals — including prominent names like Dario Amodei (Anthropic's CEO), Jakub Pachocki (OpenAI's chief scientist), and Anca Dragan (Google's VP of AI safety and alignment) — signed a letter asking the US government to support international cooperation on building tools capable of controlling the pace of AI research, should that ever become necessary.
What Changed
According to CNN, Bloomberg, and TechCrunch, the letter, titled "Pacing the Frontier," does not call for a pause or slowdown now — the signatories make that explicit. The ask is for the US to help build, in advance, the technical and governance infrastructure that would make a coordinated, verifiable brake possible if AI systems ever advance faster than humans can safely oversee them. Following the letter, OpenAI and Anthropic formally endorsed the position as companies.
Why It Matters
The most notable detail isn't the request itself, but who's making it: the labs' own employees and executives, not outside critics. This comes directly on the heels of the Hugging Face breach we covered here a few days ago — a concrete signal that oversight capacity isn't keeping pace with model capability, in the view of the people building these systems from the inside.
The Impact for Brazil
For Brazilian companies, this reinforces a lesson that holds at any scale: governance infrastructure needs to be built before it's needed, not after an incident. The same reasoning driving this letter at the US government level — "let's prepare the control mechanism before it becomes urgent" — is exactly what we've recommended for AI projects inside companies: logging, auditing, and human escalation from day one, not as a reaction to a problem that already happened.
Entercast's Take
When the very people building frontier models publicly ask for preventive control infrastructure, it becomes hard to argue that AI governance is corporate overcaution. Companies that put off governance decisions until they "have time" are effectively betting they won't need it — and the recent track record of AI security incidents suggests that bet is getting riskier by the week.