{"id":341,"kind":"story","title":"LLMs Make Fuzzing Harnesses Obsolete: Why AI Security Test Automation Unicorns Emerge","url":"https://arstechnica.com/information-technology/2026/05/mozilla-says-271-vulnerabilities-found-by-mythos-have-almost-no-false-positives/","host":"arstechnica.com","by":"mrlatte","score":1,"comments":0,"editorial_comments":0,"time":"2026-05-08T09:00:00.000Z","origin":"archive","language":{"source":"ko","reading":"en","translated":true},"summary":["Anthropic's Mythos Preview discovered 271 vulnerabilities in Firefox 150, and around the same time, XBOW became a unicorn with a $155 million Series C funding.","The investment rationale is that instead of humans writing fuzzing harnesses for each API, a bottleneck, LLMs read code and generate hypotheses to bypass this step.","However, linking reproduction, false positive filtering, and review of fixes is another product challenge, so saying harnesses will disappear is premature."],"discussion":"https://startupxo.com/en/341/","markdown":"https://startupxo.com/en/341.md","api":"https://startupxo.com/api/v1/item.json?id=341&lang=en","text":null,"edited":null,"sources":[],"thread":[]}