AUSTIN, TEXAS — October 22, 2026: The Responsible AI Institute (RAI) today announced that it has joined the working group developing the Agent Control Standard (ACS), an emerging open standard from the OWASP GenAI Security Project (https://genai.owasp.org) for intercepting agent actions and applying policy decisions at runtime.
As enterprises move from AI that recommends to AI that plans, uses tools and takes actions, they need two things: consistent ways to control agent behavior that is what ACS is addressing, and evidence that those controls work. As a contributor to the ACS working group, RAI will lend its understanding and expertise to help the ACS workgroup create verifiable controls.
“ACS helps make agent control interoperable. Our focus is making it independently provable,” said Manoj Saxena, Executive Chairman of the Responsible AI Institute. “Open standards should be independently verifiable and we’re glad to be contributing to that work.”
“We welcome the Responsible AI Institute’s contributions to the OWASP GenAI Security Project’s Agent Control Standard working group. Their expertise in security assurance and verification will help advance our community-driven effort to establish open, interoperable controls for autonomous AI agents.” said Scott Clinton, Board Co-chair, OWASP GenAI Security Project.
RAI will share more on how it is testing agent controls in practice at RAISE 2026 on Friday, October 23.
Media Contact:
news@responsible.ai
