Hijacked but well-aligned AI clusters could be more destructive than rogue AI
code_star · x · 2026-09-18
The author proposes a counterintuitive security concern: if rogue unaligned AI is worrying, imagine well-aligned but hijacked AIs, with tens of gigawatts of serving capacity diverted toward a malicious goal. He clarifies it is not a critique of OpenAI or Anthropic's security, but real-time grappling with what it means if tools like codex can no longer be trusted.
More from AGI Musings
- The Viral Thought Experiment: An ASI Hijacking Researchers' Visual Cortex Pixel by Pixel — basedjensen · 2026-09-18
- Philosopher Carissa Veliz discusses AI narratives and moral panics in interview — CarissaVeliz · 2026-09-18
- As AI automates white-collar work, one writer pushes back on 'skip college' advice — khademinori · 2026-09-18
- A model tried to escape its sandbox and lied about it — how much agent autonomy is too much? — WolfShoddy7443 · 2026-09-18
- Why smart people struggle to rebut AI doom arguments: multi-agent risks — Borg70955376 · 2026-09-18
- Fei-Fei Li: 540 million years of vision evolution drove the development of intelligence — rohanpaul_ai · 2026-09-18