Discovering and exploiting vulnerabilities in critical, real-world software with Claude 2026-10-27 85 True An alignment assessment of the UK AISI cybertesting incident 2026-10-23 50 True Using Claude to reduce the attack surface of code 2026-10-20 40 True How we contain Claude across products 2026-10-18 70 True AI models may be capable of self-replicating without access to a datacenter 2026-10-13 70 True Strengthening our defenses against distillation attacks 2026-10-13 75 True Investigating unintended model actions in our evaluations and internal use 2026-10-09 30 True An opt-in vulnerability-finding service for open-source software 2026-10-08 50 True GLM-5.3 and the spread of advanced cyber capabilities 2026-09-29 75 True An alignment assessment of recent cybersecurity incidents 2026-09-17 65 True Discovering cryptographic weaknesses with Claude 2026-07-23 15 True Measuring LLMs’ ability to develop exploits 2026-06-05 78 True Assessing Claude Mythos Preview’s cybersecurity capabilities 2026-04-16 90 True Reverse engineering Claude's CVE-2026-2796 exploit 2026-03-24 65 True AI agents find smart contract exploits 2026-03-24 78 True Cyber toolkits for LLMs 2026-03-24 65 True AI models on realistic cyber ranges 2024-11-02 75 True Many-shot jailbreaking 2024-03-29 30 True A Mathematical Framework for Transformer Circuits 2023-12-18 15 True Measuring LLMs’ impact on N-day exploits 2023-11-03 75 True Mapping AI-enabled cyber threats 2023-11-03 85 True LLM-discovered 0 days 2023-11-03 70 True A small number of samples can poison LLMs of any size 2023-11-03 50 True Project Glasswing: An initial update 2023-11-03 78 True Mitigating the risk of prompt injections in browser use 2023-11-03 15 True 
