🤝 What Happens When Attackers Can Automate Exploits? Inside Maze's AI-Native Defense
In a Theory Ventures conversation with Harry Wetherald, co-founder and CEO of portfolio company Maze, the discussion tackles how AI has turned cyberattacks into a scalable business: once an exploit works, attackers wrap it in tooling and rerun it for a few tokens, chaining together longer sequences of malicious actions. Maze counters this by running AI agents that continuously scan cloud environments for vulnerabilities, having evolved from one broad, high-freedom agent into hundreds of thousands of narrowly scoped micro-agents governed by hard permission boundaries. The hardest problem isn't raw accuracy but accuracy at an affordable price — identical runs on the same data can yield different results, and investigating millions of vulnerabilities can cost millions in inference spend. For investors, the piece signals that AI-native cybersecurity, where defenders must match attackers' automated speed and cost curve, is becoming a critical and fundable infrastructure category.
🔗 Read more