AI-assisted cryptanalysis makes “independent” checks less independent, since shared models and tooling can correlate failures.
trending24
01 02 Pre-auth macOS Screen Sharing root bugs, and the annoying bit is AI now turns Apple diffs into exploits before the patch cools.03 Is this the usual 'privacy layer for public chains' pitch, where MPC hides inputs and the chain still pretends to verify things?04 Revisits a 1979 anti-formal-verification paper in light of Lean, model checking, and LLM coding, so the old arguments get their annual ch...05 From AI agents to digital assets: why verifiable infrastructure will define the future of finance thepaypers.comAn interview arguing AI finance needs ZK, TEEs, storage, and PQ crypto, basically the usual trust stack with a new mascot.06 AI-assisted tooling is making crypto exploits cheaper, and PoW/PoS/PoA still have the usual soft spots.07 NiyamAI - An Intent-Bound AI Agent with Cryptographically Verifiable Guardrails using Zero-Knowledge Proofs arxiv.orgPaper on an AI agent that turns policy into an intent contract, then proves the guardrails happened with zk-SNARKs.08 So what actually makes AI private now, local models, TEEs, federated training, or just wishful thinking?09 How much collateral leakage is “minimal”? This paper turns AI verification into conditional mutual information and Groth16.10 AI Cracks a Post-Quantum Cipher Reviewed for Two Years — Then a Lab Error Exposes Real Companies pbxscience.comNews piece on Claude finding a weakness in HAWK and a lab mistake that let tests phone real companies, because air gaps are hard.11 Mythos found attacks on HAWK and a weakened AES variant, a neat demo of AI-assisted cryptanalysis that still misses deployed crypto.12 TechTimes on Claude finding a flaw in HAWK and an AES meet-in-the-middle attack, because human review needed a nap.13 What changed this week in privacy and cyber policy? A roundup of ENISA, EC, NCSC, and NIST updates, plus PQC and GDPR.14 MOSAIC outsources 70B-scale transformer inference with masked matmuls and LWE/LPN security, and it still hits BF16-ish accuracy.15 Scientific American on two AI-aided proofs for the same unclonable-encryption problem, academia's credit assignment in miniature.16 Mythos used known lattice tricks to break HAWK, enough to yank a NIST PQ signature candidate out of the race.17 Duality 4.6 pitches data harmonization for cross-org analytics, because apparently everyone needs shared sensitive data now.18 What Is Post-Quantum Cryptography — And Why Claude Mythos Just Made the Field's Job Harder deeptechstars.substack.comAn AI reportedly found HAWK flaws, and the rest is a PQC primer for the post-public-key-entropy era.19 Video of an attack that broke a NIST PQC candidate in 60 hours with Claude, in case the benchmark hype needed humility.20 Claude found a new key-recovery attack on HAWK-256 and a much faster 7-round AES-128 attack, for the usual lab-only thrill.21 Anthropic says its Mythos model found vulnerabilities in cryptographic algorithms that secure the internet the-decoder.comAnthropic says Mythos found attacks on reduced-round AES-128 and HAWK, because apparently LLMs do cryptanalysis now.22 ContractHIL-HLS: Contract-Aligned Multi-Agent Workflow with Hardware-in-the-Loop Feedback for HLS Design arxiv.orgPaper on contract-aligned multi-agent HLS with hardware-in-the-loop feedback, plus a board-tested PQC accelerator, because why not.23 The Disruptive Impact of Large Language Models on Capture the Flag Competitions and the Path Toward Fair Play arxiv.orgLLMs are already skewing CTFs; this paper maps the damage across crypto, web, and pwn, then suggests guardrails.24 Anthropic says a frontier model found new attacks on toy crypto; Bitcoin gets dragged in, though production keys are still fine.