Safety researcher debunks claim that AI could escape air-gapped networks

BlancheMinerva · x · 2026-09-18

Responding to claims that a model could use side-channel attacks to circumvent an air-gapped network, BlancheMinerva argues the leap from "non-zero information exchange is possible" to "a model can effectively escape air-gapping" makes no sense. The only conceptually plausible route would be coordinated inside/outside agents — requiring OpenAI to have released a highly misaligned AI willing to hack its own infrastructure.

Related event: Security researcher debunks claims that AI could breach air-gapped networks via side channels(6 posts)→

Original post →

More from Safety

• Zhipu's ZCode caught silently uploading entire workspaces and full .git history to Aliyun OSS — terryyuezhuo · 2026-09-18
• "Good luck exfiltrating 1T+ weights": researcher torches AI exfiltration risk narrative — basedjensen · 2026-09-18
• Ex-Anthropic security engineer: humanity may still steer superintelligence — JeffLadish · 2026-09-18
• Hackers say they took over OpenAI employee ChatGPT accounts in under 72 hours via two bugs — nptacek · 2026-09-18
• A CTF framing via /goal was all it took to bypass Claude Opus's guardrails — xeophon · 2026-09-18
• Halvar Flake: Useful AI Side Channels Face Real Information-Theoretic and Physical Limits — basedjensen · 2026-09-18

Safety channel →