Security researcher warns commentary is eroding trust in air-gapping at frontier labs

BlancheMinerva · x · 2026-09-19

BlancheMinerva, a longtime advocate for air-gapped networks at frontier AI labs, says she reacted strongly negatively to commentary that seemed to preemptively discredit air gapping as a worthwhile security feature.

She had hoped frontier labs were finally starting to listen to computer security experts, but observed (via a admittedly biased Twitter feed) that the net effect of these comments was to spread significant distrust rather than push real security improvements. The thread references security lessons from the OpenAI/Hugging Face hacking incident.

Related event: OpenAI's Noam Brown Says Air-Gapping May Not Stop Rogue AI, Sparking Security Debate(31 posts)→

Original post →

More from Safety

Safety channel →