Safety Evals — full archive
16 daily briefings · 2 reviews.
$ ls ai-safety-evals/reviews/ # the step-back reads
- week Control-arena v20 and inspect_evals v0.24 break APIs
- week Petri 3.1.1 closes filesystem and server-side tool escapes
- Inspect adds tool state scopes, fixes usage undercounting
- inspect_ai drops Python 3.10, grades stop failing on fenced JSON
- Docker sandbox stops unpacking agent archives on your host
- Inspect blocks writable host mounts, control-arena rescues 6 weeks of swe_django side-task scores
- Sentinel monitors gain model-call visibility, unhandled escalations now halt samples
- inspect_evals moves shared test tooling into pytest-inspect-evals 0.1.0
- Sentinel monitors gain task and sample descriptions
- Inspect evals get reproducible answer shuffling, control-arena stops running docs code
- Petri needs explicit realism roles, Opus 5.5 support lands
- Haiku 5.5 lands in inspect_ai; variadic tools stop demanding args
- inspect_ai closes sandbox bridge config leaks, inspect_evals restores official SWE-bench images
- inspect_ai recovers stalled sandbox polls, control-arena ships v20.0.0
- Inspect logs provider request ids, dataset shuffle takes an int seed
- Nova models stop breaking reasoning_effort evals
- inspect_evals 0.23.0 fixes the HLE registry lookup error
- Petri 3.1.1 closes filesystem and server-side tool escapes
Never miss a Safety Evals briefing.
Get Safety Evals's week in review every Monday - and security advisories the day they land. Free, unsubscribe in one click.
One click. Every Monday: what mattered across the projects you've been reading (or our AI picks) — security news same-day. You refine anytime. Choose specific journals · Read a sample issue →