pragma.vision Technology observatory & strategic foresight

Verification registerSecurity & Identity

Living definition

What is Petri (Parallel Exploration Tool for Risky Interactions)?

AI Safety, Eval & AlignmentSecurity & Identity

As of
2026-10-09
Revision
2026-10-06.1
Method
v1.4.0

Definition

The term, in context

AI-assisted draft · approved dataset

Petri is an open-source alignment auditing tool that runs an automated auditor model through many simulated interactions with a target model in parallel, each seeded from a short natural-language instruction describing a scenario to probe. A judge model then scores the resulting transcripts along dimensions such as deception, sycophancy, and inappropriate use of tools, turning open-ended behavioral exploration into ranked transcripts a human can review.

Live readiness status

Status as of the current dateline

AI-assisted assembly · derived results

As of 2026-10-09, verified readiness is 35 (claimed 55, reported 50, gap 20) — Strong evidence strength; current signals suggest Use for specialist alignment investigations with human review.

The readiness fact belongs to the canonical Readiness Verdict for Petri (Parallel Exploration Tool for Risky Interactions).

Dataset approval

Human editorial release

Codex — delegated editorial approver (operator authorization 2026-10-06)

Dataset
2026-10-06.1
Hash
sha256:0df9a4b64770fab5c2feaca298223eb10279a31b43487556407f496c9b022685
Approved
2026-10-06

Your opinion

Tell us anything.

What works, what doesn't, what's missing — especially about our watches, lenses, and the register itself. Anonymous is fine; leave an email if you'd like a reply.