Evidence record / eve

A dated record of one change, kept so the writing that cites it can be checked. Compiled from the sources listed below by the research run, not written for reading. The judgment lives in the signals and issues that cite it, below.

2026-09-21-eve-auto-a-model-can-now-approve-tool-calls

auto(): a model can now approve tool calls. A fourth approval helper between once() and always(): routine calls run without a human, and a classifier decides what is routine.

Channel: tagged-release. Half: capability. Date: 2026-09-17.

Operator consequence: Try it only where a wrong “clear” is recoverable. The classifier sees the tool name and input, not the downstream effect, and tool input leaves your deployment for the evaluator’s provider. It answers the contract question “does the approval-gate model change how teams gate tool calls?”: yes, eve now ships a first-party model-judged gate, which moves human attention from every call to the classifier’s false negatives.

Receipt

Finding metadata

Run: 2026-09-21-weekly-digest-2026-08-20_2026-09-21-frontier-v0

Finding ID: 2026-09-21-eve-auto-a-model-can-now-approve-tool-calls

Profile citations

  • Eve / claim / auto-approval

Source links

Primary links, including exact changelog lines when available.

Versioned source: run artifact