Institute for AI Policy and Strategy warns of frontier AI cybersecurity risks
The think tank proposes new federal coordination on AI-related cyber and national security threats.
The Institute for AI Policy and Strategy has called on the White House to develop a national security strategy focused on frontier AI systems. The recommendations follow reports concerning advanced cybersecurity-related capabilities demonstrated by Anthropic’s Mythos Preview model.
The organisation recommended stronger protection of AI model weights, expanded monitoring of adversarial AI development, and improved information sharing between government agencies and private companies. The memo also proposed enhanced testing standards and oversight mechanisms for advanced AI systems.
The organisation recommended that agencies including the NSA, ODNI, and CISA strengthen capabilities for detecting AI-enabled cyber threats and supporting defensive operations.
Additional recommendations included increased investment in AI safety research, third-party evaluations, and emergency response planning. The institute also stressed the importance of cooperation between governments, industry, and international partners in managing frontier AI risks.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!