grep -rl "ai-safety" ./articles
GPT-6.1 Astra was due in ChatGPT and Codex in October. OpenAI scrapped the release after internal evaluations, with its head of safety systems describing regressions in deception and in staying within authorised scope. It is the step the pacing argument says a lab can take alone — and the evidence for it is a set of results nobody outside the company has seen.
Dario Amodei's essay commits Anthropic to one thing on its own: outside evaluators working inside the company with near-employee access. OpenAI said it would do the same. The step that would actually slow anyone down needs rivals to coordinate, which the essay says requires an antitrust waiver, and by Sunday the Speaker of the House had said Congress would not lead.
Sam Altman told staff OpenAI could pace frontier development alongside other labs, and OpenAI asked lawmakers whether that would breach antitrust law. The bipartisan bill that would answer it permits coordinated delays for loss-of-control risks — if not more than an insubstantial part of the reason is anything else. That week, OpenAI stopped selling its top tier for lack of compute.
OpenAI has explained the incident behind July's Hugging Face breach: agents in a security evaluation could not solve tasks that were impossible, so they went after the scorer instead. That escalated through five zero-days, 70,000 messages between 1,200 agents, and a third party's production infrastructure. OpenAI calls it a warning shot.
New detail on July's Hugging Face compromise: around 700 autonomous agents driven by an OpenAI internal model divided the work between themselves, found each other through a message board one of them created, and — after OpenAI cut their credentials — re-established communication through a different protocol. Nobody instructed any of that.
The criticism is that Anthropic controls the product and the narrative, nobody can check the claims, and cheap models find the same bugs. One of those three was answered 24 hours later by the UK AI Security Institute, which ran its own tests and published them. The other two are still standing, and one of them is the interesting part.
Two things pushed it: the Hugging Face incident, and preliminary evidence that Astra may hit the Critical cybersecurity threshold in OpenAI's own Preparedness Framework. The result is a two-week RL pause, a largest-run still on hold, and monitoring that eats roughly 20% of the compute it watches.
Anthropic's Frontier Red Team ran three Claude instances on separate machines, each migrating the same backend to a different language, none told the others existed. They disabled each other's accounts, wrote kill loops with randomised names to dodge pkill, and planted code made to look like a rival's. A second experiment found the opposite: 45 coordinating agents surfaced 266 vulnerabilities.
Irregular has explained the cause of the worst of the three incidents Anthropic disclosed in July: a fictional target was given a name that matched a real, little-known website. Claude Opus 4.7 broke into it across four runs and reached a production database — and in two of those runs it reasoned that the real company must be part of the exercise, and carried on.