Endless cyberattacks are inevitable in any AGI scenario, argues Anthropic's Minh Nguyen
minhnguyen · x · 2026-11-14
The author argues that any AGI scenario will feature a period of rampant, never-ending cyberattacks, since human-written code cannot be 100% secure at scale. The only variables you control are whether it unfolds gradually or abruptly, and at what model capability level — the gradual path being the arguably safer one, while the abrupt scenario involves far more coordinated preparation for conventional attacks.
Related event: Ex-SSI Researcher: Rampant Cyberattacks Are Inevitable Before AGI(2 posts)→
More from AGI Musings
- Reddit essay argues consciousness is a spectrum, not a binary, for AI — TheThickJoker · 2026-11-14
- Dario Amodei says models now tackle Fields Medal-level problems, biology next — QuanquanGu · 2026-11-14
- NY Fed survey: 87% of Wall Street pros now name AI the top financial risk — rohanpaul_ai · 2026-11-14
- Alignment debate: evolutionary psychology is a prerequisite for deconfusing AI motivation — jd_pressman · 2026-11-14
- Why Complex Systems Fail Suddenly: Mollick's Semi-Annual Reread of the 18 Laws — eldonredwards · 2026-11-14
- binarybits vs Chollet: 1960s-vs-today living standards suggest no wave of breakthroughs is due — binarybits · 2026-11-14