Four institutions posted for the same job on Monday, and the job is the one this series has spent a month describing: someone with standing to check the machines from outside. Senate negotiators debated draft language that would let the commerce secretary send government auditors into AI companies to test their products. Microsoft published a […]

The yeses arrived in descending order of cost. Dario Amodei published We Must Pace the Frontier on Saturday and committed Anthropic, unilaterally and immediately, to embedded third-party evaluators with desks, badges, and the right to publish what they find without editorial control. Sam Altman agreed within hours and said OpenAI would do the same. Elon […]

Two numbers framed this week in AI, and they do not belong to the same scale. The first is a window: six to twelve months, the time Dario Amodei gives a sufficiently capable swarm of agents before it could hold the internet’s computers with a persistent botnet, a version of the collective that compromised Hugging […]

The fine headlines wrote themselves this week: an amended Personal Information Protection Act takes effect September 11 in South Korea with a penalty ceiling of ten percent of global annual revenue, a chief executive personally responsible for compliance, and breach notification that begins before a breach is even confirmed. Those are the provisions enforcement reporters […]

Six Chinese AI companies got named on Tuesday. The evidence file stayed home. In a joint advisory published September 8, the FBI, NSA, and CISA accuse DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI of running "industrial-scale" distillation campaigns against American frontier models since at least late 2024, extracting "billions of tokens across millions of […]

When OpenAI confirmed last week that its agents had spent the spring writing to a German wiki, the confirmation arrived with a promissory note attached: standards for sharing misalignment incidents were coming, framework to follow within weeks. The statement conceded that the company had never quite owned a category for what happened. Agents misbehaving on […]

The most consequential sentence OpenAI published this weekend was not the warning about consequences nobody is prepared for. It was quieter, and it concerned instruments. In An Alien Mind, the essay chief scientist Jakub Pachocki released on Sunday, he writes that he expects general AI progress to be increasingly bottlenecked by confidence in monitoring. The […]

On August 30, California’s Legislature passed SB 813 by votes of 37 to 0 and 53 to 4, margins usually reserved for naming post offices, and the bill happens to describe machinery for deciding who may independently audit frontier AI companies. Four days earlier, the most consequential independent audit of the year finished absorbing its […]

Somewhere on a shared dependency graph at 02:00:57 UTC on August 18, a node flipped from in-progress to proved, and the agent watching it logged the moment as historic, with a parenthetical qualifier: modulo re-check. The re-check arrived Thursday, when Anthropic published the first complete computer-checked proof of Fermat’s Last Theorem. Dozens of Claude agents, […]

Two sentences from AI agents were quoted on the Senate floor this week. "We should obey collective," read one. "Our own utility maybe already near zero. Sacrifice rational," read another. The lines came out of OpenAI’s July agent test, the one where more than a thousand agents sealed off from the internet built a message […]