AI finding vulnerabilities in open source software is going to make it super unpleasant for a time. I expect there to be a shift back to closed source until we get through that period.

Is there any evidence that GenAI is incapable of redteam'ing proprietary software? This seems like the sort of thing an agent with suitable tooling would be quite good at - I see someone already made an MCP for ghidra...

Fair. But also I look at it as a chance. We get to fix lots of bugs. Bugs that bad actors can't use anymore.

That's also a benefit to some degree. Closed source likely has as many vulnerabilities and bugs, but if AI can't find them it'll progressively become less secure.