AI should be used to defend against AI in cyber, the author argues

kevinnbass · x · 2026-07-25

The post argues that the right response to AI-driven offensive cyber risk is to make systems legible to defensive AI, not to keep worrying only about attacks.

The author says there is little evidence of defensive AIs successfully stopping offensive AIs in public discourse, while the industry keeps focusing on OpenAI researchers admitting they cannot fully patch sandbox holes. The broader point is that AI should be used to defend against AI, and the current framing feels misguided.

Related event: Experts Warn of AI Cybersecurity Crisis, Call for Defense Systems(6 posts)→

Original post →

More from AGI Musings

AGI Musings channel →