GOVERNATLAS / PRODUCT DOSSIER

Inspect AI

Inspect documents an open-source evaluation framework with datasets, solvers, tools, scorers and evaluation logs, developed by the UK AI Security Institute and Meridian Labs.

Explore the product ↗Add to comparison ↗

01 / CONTEXT & SCOPE

Understand the proposition.
Define the boundary.

The decision to make

Distinguish the evaluation task from the model and its tools. Dataset, solver, scorer and sandbox configuration jointly determine what an agent can attempt and what the evaluation can observe. An official research framework is not approval of a specific product or permission to expose a real system to generated actions.

What this record does not establish

Developed by the UK AI Security Institute and Meridian Labs. Framework availability, prebuilt tasks and sandbox support do not establish government endorsement, complete isolation or enterprise application safety.

1 current source · 1 published assertion
Source coverage is a research boundary, not a quality score.

02 / YOUR EVALUATION

Turn interest into a useful trial.

Open the evaluation playbook ↗

A test worth running

Build a synthetic task with a mock read-only tool, a scorer that checks a concrete expected outcome and one deliberately unobservable result. Compare two configurations using retained logs. Review the sandbox permissions and explain which failures arose from the task runner rather than model behavior.

GovernAtlas editorial proposal. We have not performed this test.

Evidence to bring back

A dataset/solver/scorer/sandbox manifest with versioned permissions.
An evaluation log distinguishing behavioral failure from execution failure.

03 / DOCUMENTED EVIDENCE

Read the claim. Inspect the basis.

Each statement retains its classification and preserved source. A vendor assertion is not an independent test result.

first party asserted

Inspect documents an open-source evaluation framework with datasets, solvers, tools, scorers and evaluation logs, developed by the UK AI Security Institute and Meridian Labs. Scope: primary documentation only; no independent product test, evaluator accuracy or compliance determination.

Source, dates & preserved provenance

Inspect AI — primary scope and current identity ↗

Authority: first party
Observed 2026-10-16 · Reviewed 2026-10-16 · current until 2026-11-15

Artifact SHA-256: 2603bc1fd064b1dada8ab9d1cb09c519b3be8db37de594baedaf6d027312ffd7

The hash identifies the preserved capture. The live source may have changed.

Source register

Inspect AI — primary scope and current identity ↗first party · captured 2026-10-16

Brand asset provenance

Organization identity asset from its official website; inclusion does not imply endorsement.

Official website ↗ · Original asset ↗

04 / BEFORE YOU COMMIT

Better questions.
A clearer decision.

Resolve these with the provider and your own evaluation team.

Can you provide a dataset/solver/scorer/sandbox manifest with versioned permissions for the selected configuration?

Can you provide an evaluation log distinguishing behavioral failure from execution failure for the selected configuration?

Which exact version, entitlements, data paths and responsible roles apply?

Regulatory applicability requires a separate, current determination. Inclusion is not certification.

KEEP THE RECORD USEFUL

A dossier should stand up to scrutiny.

Have a source that changes this picture? Submit it for editorial review.

Raise an evidence issue ↗