Open-source OWASP security testing for AI models and agents

1 pointsposted 3 hours ago
by alysheaib

1 Comments

alysheaib

3 hours ago

We built this after Promptfoo got acquired by OpenAI (March 2026). The AI security testing space needed a vendor-neutral alternative — especially with the EU AI Act deadline hitting August 2.

  Tessera runs 42 automated tests across 5 OWASP categories: model security, application security, infrastructure, data governance, and agentic
  AI. It's the first framework with complete OWASP Agentic AI Top 10 (ASI 2026) coverage — 10 tests for goal hijacking, tool misuse, rogue agents,
   cascading failures, etc.