MITRE has launched a new AI Incident Sharing initiative aimed at enhancing community knowledge of AI threats and defenses.

This initiative involves collaboration with over 15 organizations, including major companies like Microsoft, Citigroup, and CrowdStrike.

The initiative expands existing incident sharing efforts by introducing new generative AI attack techniques and mitigation methods.

Its goal is to create a safe environment for distributing sanitized AI incident information, thereby improving collective defense against threats.

Douglas Robbins, VP of MITRE Labs, emphasized that rapid and standardized information sharing will enhance the collective defense of AI systems.

Participants in the initiative will receive protected and anonymized data on real-world AI incidents, with organizations able to submit incidents via a dedicated web portal.

The initiative utilizes the STIX data schema for traditional intelligence sharing, allowing organizations to submit incident data for consideration in a trusted community.

This effort is part of the Secure AI project, which expands the MITRE ATLAS community knowledge base that has been gathering anonymized incident data for two years.

In November 2023, MITRE collaborated with Microsoft to update the ATLAS framework to specifically address generative AI threats.

The Secure AI project has enhanced the ATLAS threat framework with new case studies and mitigation strategies focused on generative AI.

MITRE has a similar initiative in aviation safety through the Aviation Safety Information Analysis and Sharing database, demonstrating its commitment to secure information sharing.

Overall, the initiative aims to facilitate secure sharing of information regarding threats, attacks, and incidents related to AI-enabled systems.