30 September 2026, San Marino – The Parliamentary Assembly of the Mediterranean (PAM), supported by its Centre for Global Studies (PAM-CGS), just published a new report examining AI-enabled cyber threats, the growing capabilities of autonomous systems and their implications for national and international security. The publication, available here, supports legislators, policymakers and technical experts in assessing how responsibility, oversight and intervention should evolve as AI becomes more powerful and widely accessible.
The report is up to date with current debate as of 29 September, and systematic integrations will be disseminated regularly.
The report identifies two increasingly interconnected risk pathways: human-directed misuse, involving malicious actors employing AI to support deception, cyber intrusion, malware development and weapons-related research; and unauthorized agent behavior, whereby systems pursue assigned objectives through actions exceeding the permissions or boundaries established by their operators. It also examines how connected digital activities may contribute to physical harm, including risks affecting critical infrastructure and dual-use biological research.
A central conclusion is that effective human control requires adequate institutional capacity. Those responsible for supervision must have sufficient information, expertise, authority and technical means to assess and interrupt consequential actions.
Among its recommendations, the report calls for documented assessments of high-risk AI deployments, independent expertise where potential consequences are serious, and timely reporting of significant incidents to competent authorities. It recommends clearly allocating responsibilities among developers, deploying organizations and infrastructure operators, with obligations covering security, evidence preservation and cooperation during investigations.
The report also recommends safeguards for AI-assisted parliamentary work, including source verification, protection of confidential information and clearly assigned human responsibility for official texts.
For the Euro-Mediterranean and Gulf regions, the report identifies opportunities for cooperation on common definitions of serious incidents, evaluation criteria, evidence preservation and cross-border notification. Shared expertise and training could strengthen preparedness in countries with limited technical resources.
PAM-CGS, with the support of its Global Permanent Parliamentary Observatory on the Misuse of AI and ICT, will continue to monitor emerging risks and provide independent research to inform its international network.//