Zama founder: AI can't break FHE, and even if it did, funds stay safe
juanbenet · x · 2026-10-10
Responding to questions about AI's impact on FHE, Zama's founder laid out four points:
- No evidence that AI, SI, or quantum computing can break lattices/FHE — full lattice breakage might even imply P = NP.
- Better attacks will almost certainly emerge, requiring larger parameters and a short-term performance hit; Zama's modular protocol can swap components without UX changes or key rotation.
- Zama uses FHE only for privacy: breaking it would expose balances but not enable theft of shielded funds — unlike ZK, where breaking the crypto means stealing funds and minting tokens.
- The team invests heavily in cryptanalysis to improve FHE security.
More from Safety
- Vanity Fair: ChatGPT helped write my daughter's suicide note, and Altman won't release the logs — kyliebytes · 2026-10-10
- Subliminal learning doesn't transfer across model families, narrowing alignment propagation concerns — Miles_Brundage · 2026-10-10
- NYU study: incidental chatter contaminates 35% of LLM clinical notes and disrupts reasoning — NYU-OLAB · 2026-10-10
- Aaron Grattafiori explains how LLMs hunt patch variants at scale — dyn___ · 2026-10-10
- Seven national blue teams ran THOR scanner at Locked Shields 2026, want LLM-assisted analysis — cyb3rops · 2026-10-10
- Are AI companies faking doomer fears to distract from real harms? Vox says probably not — Miles_Brundage · 2026-10-10