$ ls ai-safety-evals/reviews/ # the step-back reads
Safety Evals in review
One review per week (and month): what shipped, what broke, what to act on.
Oct 5–11, 2026 weekly [act] 8 stories
Control-arena v20 and inspect_evals v0.24 break APIs
Both releases ship breaking changes; inspect_ai adds Claude Haiku 5.5 and Claude Opus 5.5 support and provider request id logging.
Sep 28–Oct 4, 2026 weekly [act] 8 stories
Petri 3.1.1 closes filesystem and server-side tool escapes
The release also drops target tool options that let model providers run tools inside the API call, where the auditor never sees them.
Keep up with Safety Evals in about 3 minutes: what actually shipped — the commits, pull requests, releases, and security advisories that matter.
fair warning: these emails are deeply technical. diffs, version numbers, CVEs, benchmark deltas. if that's not your idea of a good read, this isn't your newsletter.
Folds into your digest — weekly by default, monthly if you prefer. Unsubscribe in one click.