Claude Fable breaks top SMHasher hash functions in a day, security drops from 64 bits to zero
thomasahle · x · 2026-09-20
Researcher thomasahle (Normal Computing) reports that Claude Fable found collisions in komihash, a5hash, HighwayHash, SpookyHash, aHash, and t1ha2 within a single day, with several top SMHasher entries dropping from "64-bit security" to "at most 32 bits" or zero.
- Why it matters: universal hashing guarantees low collision probability over a secret key (e.g. 2^-64) and underpins hash tables and load balancing; heuristic hashes have inputs that collide far more often, enabling load-skew and DoS attacks. Adversarial inputs used to be hard to find — agent swarms change that.
- Author's view: a fascinating post-AI domain — agent swarms can hunt vulnerabilities in weak crypto, while AI is strong enough to help prove hardness guarantees for good crypto.
- Recommendation: no need to switch everywhere to slow cryptographic hashes like SipHash; use provably safe universal hash functions, some with Lean proofs.
Related event: Researchers use Claude to crack six major hash functions in a day(3 posts)→
More from Safety
- Senate Bill Would Force Big Platforms to Accept AI Agents, Raising Auth Privacy Questions — Prof_CarlaReyes · 2026-11-12
- United Airlines Bans AI Agents From Booking Flights Over Data and Legal Risks — rohanpaul_ai · 2026-11-12
- Stanford HAI blueprint: 56 US agencies report 3,611 AI use cases, evaluation lags — StanfordHAI · 2026-11-12
- Focused task framing boosts agent boundary testing by ~60 points; VM escape still fails — perplexity_ai · 2026-11-12
- Perplexity red-teams its SPACE VM sandbox: CTF prompts mislead agent security testing — perplexity_ai · 2026-11-12
- Narayanan: blanket bans on teen AI companions are bad policy; incentivize pro-social design instead — random_walker · 2026-11-12