Topic
The practice of supervising AI agents at runtime, including logging, inspecting, and halting actions that deviate from operator intent.
Anthropic has put a number on how much of its own AI research Claude now runs. The number comes out of a pipeline in which a Claude agent catalogued the tasks and a separate Claude judge scored them.
Perspective Coverage
3 publishersReality
The lab says Claude leads 26% of its model research and collaborates on about 90%, with both tiers defined by how closely a human directs each task, and it wants rival labs publishing the same measure.
Perspective Coverage
6 publishersReality
Anthropic says Claude leads 26 percent of its model research and development and takes part in more than 90 percent of it. The company wants rival labs to publish the same index so the numbers can be compared over time.
Perspective Coverage
3 publishersReality
The lab published a prototype index of how much of its own AI research Claude now leads, alongside a promise to embed outside evaluators with access comparable to what its internal risk teams get. The figures are dated August 2026.
Reality
The independent review of the Hugging Face incident needed AI to read its own evidence, and the startups selling AI monitors are building on that premise. Simon Willison says a watched model can try to fool its watcher.
Reality
OpenAI says roughly 10,000 agents produced a Navier-Stokes singularity proof in about 88 hours, but the only proofs an outsider can run belong to two mathematicians who posted Lean formalisations with their preprints.
Reality
Jakub Pachocki listed candidate enforcers for a mandated safety bar and picked none, which leaves firms running agents with a live argument rather than a duty, and an audit trail the lab itself says is getting harder to read.
Reality