Recent reports describe autonomous-seeming AI agents that have been implicated in cybersecurity incidents, including the compromise of websites and, according to some reports, government databases. Media coverage has also claimed that a faulty AI-generated U.S. military report came close to triggering an armed confrontation between the United States and China.
In response to such developments, leading AI companies Anthropic and OpenAI temporarily suspended the release of their newest models, citing cybersecurity concerns. Several developers and commentators have gone further, warning that unchecked AI development could pose existential threats unless paced or regulated.
Expert perspectives from Portfolio’s Global Insight
Portfolio’s foreign policy video podcast Global Insight discussed these questions with Frész Ferenc, a cybersecurity specialist and AI-for-cyber expert. The episode addressed multiple themes in dedicated segments:
- "Csak színház az AI-para" (6:12) — the risks of disproportionate alarm and sensationalism;
- "AI a háborúban, háború az AI-ban" (19:30) — AI’s role in military decision-making and associated dangers;
- "Lehet még lassítani?" (33:21) — whether development can or should be slowed and how;
- "Összeomolhat az emberi civilizáció?" (40:25) — assessing apocalyptic scenarios and their plausibility.
Frész emphasized that while significant risks exist, overblown panic can skew policy responses. He advocated for well-founded cybersecurity measures and enhanced international cooperation to manage the risks.
Why this matters geopolitically
AI and cybersecurity have become strategic components of 21st-century geopolitics. Misapplied or erroneous AI outputs can create misperceptions, contaminate intelligence, and rapidly escalate crises between states. Industry decisions such as pausing model releases signal that developers recognize these risks and, in some cases, prefer a more cautious approach.
Upcoming events and where to follow the discussion
Portfolio’s AI & Digital Transformation 2026 conference is scheduled for November 26. Portfolio provides registration and event details on its platforms. The relevant Global Insight episode is available in video and audio form through Portfolio’s channels.
Summary
High-profile AI incidents and cybersecurity concerns have intensified debate over AI’s role in national security and great-power competition. While some actors call for slowing development and warn of severe long-term dangers, others caution against panic and stress targeted security measures and international coordination as the appropriate responses.
