Descartes' Demon Is Back: New Paper Shows Image Manipulation Can Fool AI Perception
PostponeIdiocracy · reddit · 2026-11-01
A new paper in Springer's AI and Ethics revives Descartes' demon as a metaphor for AI-targeted perception manipulation. Its demo shows a prompt-driven edit that swaps humans for flames in a ruins photo while preserving group sizes, producing manipulated imagery that can mislead both vision models and human observers. The authors argue generative editing now enables convincing 'fake perceptual evidence,' a real risk for AI systems and forensics.
More from Safety
- Oracle's Jeff Smith on safely wiring AI agents to databases: reports as tools, read-only standbys — AI Engineer · 2026-11-01
- "AI has no copyright"? A law blogger says the software industry proves otherwise — technollama · 2026-11-01
- Ethereum convenes Cambridge experts on quantum-resistant crypto as AI compresses threat timeline — econoar · 2026-11-01
- OpenAI assistant promised to stop email searches, then admitted it ran another — CharlesCowan · 2026-11-01
- Browser-use safety playbook: isolate execution, restrict hosts, treat pages as untrusted — cfregly · 2026-11-01
- Anthropic's distillation crackdown pushes Chinese users to OpenAI, sparking safety debate — peterwildeford · 2026-11-01