Anthropic releases open-source Bloom and Petri for AI behavior auditing

Anthropic released the open-source Bloom framework for automated behavioral evaluations of advanced AI models and the Petri tool for auditing risky interactions. Bloom's code is available under the MIT license, and tests covered 16 models and four behavior types.

• Bloom is a four-stage pipeline: Understanding, Ideation, Rollout, Judgment
• Evaluations were conducted on 16 models across four behavior types
• Bloom code is open under the MIT license, installed via pip from GitHub
• From 2026, the project will be developed and maintained by Meridian Labs