OpenAI Cyber Summit, London, September 2026

The Defender's Window

In September, I spoke to 150 cybersecurity leaders from across Europe at the OpenAI Cyber Summit in London. The window in the title is the time defenders have to use AI to find and fix weaknesses before attackers can exploit them. It won’t stay open indefinitely.

The OpenAI Cyber Summit stage. Watch the full keynote on YouTube.

From findings to fixes

When AI makes it easier to find vulnerabilities, the bottleneck moves. Someone still has to reproduce each issue, work out whether it’s exploitable, and test the patch. I talked about building a defense factory: a repeatable process that gives the people responsible for a codebase evidence they can act on and a tested fix they can review.

Start with one repository. Give the agent an environment where it can build the code and run the tests. Once the process produces fixes your team can trust, run it regularly in CI and expand to more repositories.

The summit also covered Daybreak for Frontline Defenders, OpenAI’s $1 billion commitment to subsidised access and support for organisations that protect essential services, and the Daybreak Defense Network.

Original LinkedIn post