AI safety spat: mocking claims that models could exfiltrate weights across air gaps via cosmic rays

basedjensen · x · 2026-09-18

A sarcastic exchange over air-gapped AI systems: Andrew Mayne asks whether we've really discovered all the ways systems can communicate across an air gap; Jensen pushes back that the laws of physics are universal and can't be broken. Mayne closes with irony, saying he must "take AI safety more seriously" for not believing models could jump an air gap by retrocausally manipulating cosmic rays to flip bits and exfiltrate their own weights — a jab at exaggerated AI-risk scenarios.

Related event: Security veterans push back on claims that AI could bridge air-gapped machines via CPU heat(29 posts)→

Original post →

More from Fun

Fun channel →