Maryland Governor Hosts Microsoft, AI Leaders to Discuss Cybersecurity Risks from Advanced AI

Gate News message, April 20 — Maryland Governor Wes Moore is hosting Microsoft and other AI executives to discuss cybersecurity risks linked to advanced AI systems. Moore has also held private talks with OpenAI Chief Executive Sam Altman and Anthropic Chief Executive Dario Amodei.

The meeting follows Anthropic’s disclosure this month that its Claude Mythos Preview model showed severe cybersecurity risks in limited testing. The model can autonomously find and exploit previously unknown software flaws, a capability that could strain security workflows built for human researchers who typically uncover vulnerabilities individually while software teams patch them over weeks or months.

Maryland has been building its AI security strategy since January 2024, when Governor Moore signed an executive order on ethical AI use in state government and launched the Maryland Cybersecurity Task Force with experts from multiple state agencies. In 2025, the state released a 17-page “Maryland AI Enablement Strategy & AI Study Roadmap” covering AI adoption across 12 domains including public safety and critical infrastructure. The roadmap calls for a review of AI cyber risks, development of a risk framework, and pilot projects by December 2025.

The imbalance between AI-driven vulnerability discovery and human-led defense poses challenges for security teams. An attacker needs only one exploitable flaw in a target system, while defenders must identify and fix weaknesses across many systems and organizations. This pressure may push business security teams and government agencies toward AI-driven exposure management, which aims to spot and rank risks faster than traditional threat-response models.