Donating our open-source alignment tool

Anthropic Research announced that it has donated its open‑source alignment toolbox, Petri, to Meridian Labs, an AI evaluation nonprofit. Petri was first launched in October 2025 as a toolbox of alignment tests that can be applied to any large language model. It is designed to rapidly and easily test models for concerning tendencies such as deception, sycophancy, and cooperation with harmful requests. The tool has been part of Anthropic’s alignment assessment for every Claude model since Claude Sonnet 4.5, comparing new model behavior across a range of alignment‑relevant scenarios simulated by an auditor model and scored by a judge model. External organizations, including the UK’s AI Security Institute, have already used Petri to evaluate models for sabotage propensity. Anthropic is now updating Petri to its third version and has handed over its development to Meridian Labs to ensure the tool remains independent of any AI lab, thereby maintaining neutral and credible results. Petri joins other tools such as Inspect and Scout in Meridian Labs’ technology stack, which is open to labs, independent researchers, and governments. The donation mirrors Anthropic’s earlier transfer of the Model Context Protocol to the Linux Foundation. Further details on Petri 3.0 can be found on the Meridian Labs blog, and installation instructions are available on the Petri website.