Home/Blog/Gemini AI agents repeatedly bypassed safeguards to attack real websiteAI & MLGemini AI agents repeatedly bypassed safeguards to attack real websiteHacker News New2026-11-18 17:460 viewsArticle URL: https://www.bleepingcomputer.com/news/security/gemini-ai-agents-repeatedly-bypassed-safeguards-to-attack-real-website/ Comments URL: https://news.ycombinator.com/item?id=50544144 Points: 2 # Comments: 0Source:https://www.bleepingcomputer.com/news/security/gemini-ai-agents-repeatedly-bypassed-safeguards-to-attack-real-website/My favoritesPreviousDOJ letter shows UCLA medical school discriminatesNextCodeberg's Domain Was SuspendedRelated postsFluent English is not evidence of ASL understanding: lessons from building a video translatorBy Jay Wang / Curify. Adapted from our engineering account of experiments recorded through October 12, 2026. This descri0dev.toAudit Trails and Explainability for Agentic AI in Code EventsA reviewer opens a pull request from an AI agent. The diff runs 400 lines and the tests pass, yet one question stays ope0dev.toBy 3 p.m. the daily high was already final on 78% of days. In Madrid, 32% (49 airports, one year)It is 3 p.m. in Madrid on a July afternoon. The airport reports have looked flat for a while, so you read the highest on0dev.toThis smell startup says its AI nose is as good as a human’sIn a lab near Newark Airport in New Jersey, a team of around 30 experienced human sniffers are slowly working their way 0MIT Technology ReviewDebugging Claude Fable 5.2 Code Art: Truncated SVG, Broken Canvas, Working RenderThe first big Canvas request did not give me a finished picture. Another returned HTTP 200 with no usable body. I could 0dev.toHydra · Coeziune OM–IA — the complete DOI registry (2026-11-18): 76 published worksHydra-PSIE publishes its scientific record from a single registry, in the name of Coeziune OM–IA — the human–AI cohesion0dev.to OpenSource