Covert AI Exfiltration Channels Are Impractical: Under 10 GB/Week, Need Dedicated Hardware
BlancheMinerva · x · 2026-09-18
The author pushes back technically on the idea that AI models could secretly exfiltrate information via covert physical channels:
- Would a human or AI inside the OpenAI building take instructions from the model over a radio antenna? Every physical covert channel mechanism she has seen requires sensors and dedicated hardware.
- Even if you solve the physics issues and make major breakthroughs, the result is a one-way channel of under 10 GB/week to an observer with internet access.
A pragmatic rebuttal to AI-covert-communication safety worries: prohibitively costly and extremely low bandwidth.
More from Safety
- Cryptographers push back: claims ECC is likely to break are not consensus — jedisct1 · 2026-11-02
- UK's 'AI sovereignty' push undermined by its own energy and copyright policy — NinaDSchick · 2026-11-02
- Auditing my MCP credential broker: the worst hole was the config file — Relevant_Mousse_9117 · 2026-11-02
- Cambricon co-founder urges shared US-China AI ethics standards for leading models — pstAsiatech · 2026-11-02
- Vitalik's "post-AI cryptography": pseudonyms are dead, per-action anonymity is in — devanshmehta · 2026-11-02
- 6 prompts to see what ChatGPT quietly knows about you — and wipe it — Aiden_Tech_Ai · 2026-11-02