Newcomer says cybersecurity looms larger than existential AI risk for startups

Newcomer says founders it spoke with were focused on cybersecurity amid the AI safety debate. It cites an Anthropic threat intelligence report detailing bad actors’ efforts to misuse Claude and says leading AI models are very good at finding, patching or exploiting computer vulnerabilities. For people outside frontier model development or high-level policy, Newcomer describes cybersecurity as a practical way to think about AI safety.