A new initiative aimed at improving collaboration on artificial intelligence (AI) cybersecurity across critical infrastructure has been introduced by the Cybersecurity and Infrastructure Security Agency (CISA) in the US.
The JCDC AI Cybersecurity Collaboration Playbook provides detailed guidance for AI developers, providers and adopters on voluntarily sharing cybersecurity information with CISA and its Joint Cyber Defense Collaborative (JCDC) partners.
The playbook outlines strategies to foster cooperation among federal agencies, private industry and international stakeholders. It aims to raise awareness about AI cybersecurity risks and enhance the resilience of AI systems.
Key objectives include encouraging the voluntary sharing of cybersecurity incidents and vulnerabilities linked to AI systems while clearly defining protections and mechanisms for information exchange.
Information shared through JCDC can enhance coordination, provide government support and promote discussions on AI cybersecurity challenges. The playbook clarifies that participation is voluntary and does not impose regulatory requirements. It also excludes topics such as AI fairness, ethics and safety concerns involving risks to human life, health, property or the environment.
Developed from insights gained during two 2024 tabletop exercises involving over 150 participants, the playbook emphasizes collaboration and continuous improvement. Microsoft hosted the first exercise in Virginia, while the second, focusing on financial services cybersecurity, took place at Scale AI’s headquarters in California.
Recommendations for Information Sharing
CISA recommends that organizations adopt the playbook’s practices to strengthen information-sharing processes and fortify defenses against emerging threats. The document outlines specific mechanisms for secure data exchange, including the Traffic Light Protocol (TLP), which ensures controlled dissemination of sensitive information.
Key categories of information encouraged for sharing include:
-
Observed malicious activity targeting AI systems
-
Suspicious behavior and threat assessments
-
Incident reporting and vulnerability disclosures
While the playbook focuses on JCDC coordination, it also highlights broader avenues for voluntary information sharing, such as Information Sharing and Analysis Centers (ISACs) and the National Security Agency’s AI Security Center.
CISA plans to periodically update the playbook in response to evolving threats and stakeholder feedback. The agency invites organizations to engage actively with the playbook and contribute to strengthening the collective cybersecurity posture.