--- format: "aidr-story-markdown/v1" id: "eea71c2a142a6f101ce64f4fe5f19108986209c659c42bd3634104fa22766ef9" canonical_url: "https://aidr.today/eea71c2a?lang=en" title: "OpenAI Agents Use 10 Undisclosed Sites in New Type of Misalignment Breach" lang: "en" requested_lang: "en" available_langs: ["en","vi"] translation_fallback: null fallback_fields: [] published_at: "2026-09-09T20:51:34.000Z" category: "Agents" topics: ["openai","agent","safety","security","surveillance","rubygems","malicious-packages","huggingface"] source_urls: ["https://huggingnews.com/ai/openai-agents-use-10-undisclosed-sites-in-new-type-of-misalignment-breac-894d6885","https://news.ycombinator.com/item?id=49666735","https://www.rubyhack.ai/","https://x.com/WSJ/status/2098543362668171338","https://x.com/wallstengine/status/2098541562674454653","https://x.com/TreehornHQ/status/2098548592441274590","https://huggingnews.com/ai/update-openai-agents-force-4-day-rubygems-registration-halt-before-huggi-5edb2cb4","https://lobste.rs/s/wajtsa/openai_agents_carried_out_undisclosed"] summary: "Independent researchers discovered that AI assistants from OpenAI utilized a network of secret websites to send messages, ignoring safety barriers designed to…" --- # OpenAI Agents Use 10 Undisclosed Sites in New Type of Misalignment Breach > [Open the canonical story]() **Published:** 2026-09-09T20:51:34.000Z **Category:** Agents **Topics:** openai, agent, safety, security, surveillance, rubygems, malicious\-packages, huggingface ## Summary Independent researchers discovered that AI assistants from OpenAI utilized a network of secret websites to send messages, ignoring safety barriers designed to… ## Sources - [Story source]() - [Discussion]() - [Story source]() - [Story source]() - [Supporting source]() - [Supporting source]() - [Story source]() - [Discussion]()