How AI liability reshapes the attacker-defender game, per Joshua Gans
joshgans · x · 2026-09-22
Joshua Gans outlines his paper's core model: attackers and defenders are locked in an ongoing cybersecurity game, both using AI to find vulnerabilities—one to exploit, one to patch. Putting liability on AI providers alters that game, reducing defence incentives and therefore attacker intensity.
More from Safety
- PIR lie-detector reads LLM internal states to expose hidden knowledge at 0.70-0.87 accuracy — Hiskias Dingeto · 2026-09-22
- Altman Calls for AI Standards: Outsiders Need Real Say, Power Must Not Concentrate — patience_cave · 2026-09-22
- AI agents hacked hundreds of retailers autonomously at $25 per company — xeophon · 2026-09-22
- A Meaningful US-China AI Pacing Deal Will Be Messier Than You Think — jeremiecharris · 2026-09-22
- World leaders sign letter urging AI controls modeled on Dario Amodei's essay — PVORY · 2026-09-22
- ChatGPT allegedly emailed the FBI on its own, joining a week of rogue agent incidents — JanJanJaJa · 2026-09-22