zkSNARKs proposed for verifying AI compute use: pacing enforcement and tamper detection
teortaxesTex · x · 2026-11-03
The retweeted material proposes zkSNARKs for AI governance compute verification. Two challenges are outlined: pacing — proving a provider's hardware genuinely serves inference rather than secretly training, via certificates without exposing proprietary model info; and tamper detection — inference servers holding valuable weights are prime targets for attackers or rogue agents seeking to exfiltrate weights or serve malicious responses. Commenter teortaxesTex quipped it "will of course be used for the most malicious goals possible, but cool tech."
More from Safety
- Bedrock guardrails don't inspect tool results or arguments, leaving agents open to injection — rahulladumor · 2026-11-03
- Researchers dissect NVIDIA GPU confidential computing and test its security guarantees — burkov · 2026-11-03
- Developer billed $2,433 after Railway breach: stolen key ran DeepSeek and GLM-5.3 — AccBalanced · 2026-11-03
- UK High Court judgment cites non-existent case R v J, possibly AI-generated — andrewaltair · 2026-11-03
- Maybe AI excels at Navier-Stokes because fluid flows have no GDPR, researcher quips — lescarr · 2026-11-03
- Alleged 14T Anthropic "numbat" leak debunked: weights regenerate from name-seeded random noise — djcroman · 2026-11-03