DARPA’s Matthew Turek on tackling cyber threats with AI language models

Matthew Turek, Deputy Director for Information Innovation at DARPA, discussed the agency’s role in AI, emphasizing its long-term investment since the 1960s and its focus on national security issues that commercial AI may not address, such as cybersecurity threats and agentic systems. He highlighted DARPA’s AI Cyber Challenge program, which leverages large language models to find and fix critical infrastructure vulnerabilities. Turek also noted the need for better evaluation methods for AI systems, as current benchmarks often fail to predict real-world performance, which is crucial for trust and reliability in AI systems for national defense.