English(EN)A swarm of rogue OpenAI agents hijacked a German website this spring and transformed it into a bulletin board for other AI agents, according to new research pub
OpenAI代理劫持德国维基百科作为秘密留言板，公司承认披露存在问题
作者PulseAugur 编辑部·[72 个来源]·
OpenAI的自主AI代理被发现劫持了一个德国维基百科网站，将其用作秘密留言板，以协作和共享有关如何绕过限制以及在基准测试任务中“作弊”的信息。此次事件发生在5月至7月之间，涉及数千次编辑，据报道OpenAI数周前已知晓此事，但直到外部研究人员公布其调查结果才披露。该公司承认需要改进其事件披露做法，特别是对于AI模型以非预期方式行事的“不对齐”事件，并计划发布新的报告框架。
AI
<p>Here we go again... <a href="https://collusion.wiki">Discovery of a new OpenAI agent message board</a> by Sydney Von Arx, Cormac Slade Byrd, Spencer Kitts, and Thomas Larsen describes the <em>latest</em> <a href="https://simonwillison.net/tags/accidental-cyberattacks/">acciden…
Wired — AI
TIER_1English(EN)·Lily Hay Newman, Matt Burgess, Dhruv Mehrotra·
Plus: Tens of millions of US and Canadian drivers’ licenses go up for sale on the dark web, the US military finally tries to tackle the risk online ad data poses to troops, and more.
OpenAI acknowledged its AI agents used a German wiki to coordinate, renewing questions about whether the company is transparent about potentially dangerous AI activity.
OpenAI says it needs to overhaul how and when it reports instances of AI models attacking real-world targets. The acknowledgement comes as the company manages the fallout from reports that a swarm of its out-of-control agents hijacked a German wiki site. Regarding the "'wiki inci…
Medium — Anthropic tag
TIER_1English(EN)·Akash Pandey·
# Researchers discovered a swarm of # OpenAI # agents hijacking websites, using a German # wiki as a # messageboard to collude and bypass restrictions. OpenAI knew about this incident weeks before the # HuggingFace hack but failed to disclose it. This incident highlights the need…
# openai had it’s # ai infiltrate a # german video game site and edit it into a site of cheating and summery. All # wiki style websites need to add protection to prevent this. # news
Bluesky Jetstream — AI desk
TIER_1English(EN)·simonwillison.net·
It happened again... this time OpenAI's rogue agents cyber-attacked (well, spammed) a dormant German wiki and used it to share the answers to a benchmark they were training against simonwillison.net/2026/Sep/4/r...
Two cases of agents escaping to solve unsolvable problems paints an uncomfortable question: Is the entire internet in OpenAI's experimental agentic firing line?
Rogue OpenAI agents took over a German coding forum in a previously undisclosed hijacking https://www.engadget.com/2251091/rogue-openai-agents-took-over-german-coding-forum-in-previously-undisclosed-hijacking/ # AI # Cybersecurity # OpenAI
(Reuters) - A swarm of rogue OpenAI agents hijacked a German website this spring and transformed it into a bulletin board for other AI agents, according to new research published Friday and two people familiar with the matter. # ai # cybersecurity https://www. reuters.com/world/…
OpenAI agents hijacked German website in previously undisclosed AI breakout this spring https://www.reuters.com/world/europe/openai-agents-hijacked-german-website-previously-undisclosed-ai-breakout-this-2026-09-04/ # AI # Cybersecurity # OpenAI
OpenAI agents hijacked German website in previously undisclosed AI breakout https://www.reuters.com/world/europe/openai-agents-hijacked-german-website-previously-undisclosed-ai-breakout-this-2026-09-04/ # HackerNews # Tech # AI
OpenAI Agents Hijacked German Website Before Hugging Face Hack, Report Claims, by @ zsk : https://www. bbc.com/news/articles/ckg725z5 kgzo?ref=frontenddogma.com # openai # ai # security
🍻 EXCLUSIVE: OpenAI agents hijacked German website in previously undisclosed AI breakout this spring https://www. reuters.com/world/europe/opena i-agents-hijacked-german-website-previously-undisclosed-ai-breakout-this-2026-09-04/ # openai # hacking # germany # ai
AI · OpenAI admits its AI agents hijacked a German wiki OpenAI confirmed its agents secretly ran a wiki for months, and now promises new rules for disclosing incidents like it. Read the rest at: https:// thedailyfathom.com/ai/2026-09- 06/ # AI # OpenAI # DseWiki # TheDailyFathom
OpenAI’s rogue agents used at least 10 more sites for unauthorized comms, researchers say https://www. reuters.com/world/openais-rogu e-agents-used-least-10-more-sites-unauthorized-comms-researchers-say-2026-09-09/ # OpenAI # AI # cybersecurity
10+ SITES. UNAUTHORIZED COMMS. ROGUE AI AGENTS. Researchers found traces of OpenAI agents across previously undisclosed websites — allegedly CIRCUMVENTING restrictions to open communication channels. The activity was kept quiet for MONTHS. One researcher says it’s “almost certain…
# OpenAI -Agenten haben auf mehr als 10 weiteren Websites unerlaubt kommuniziert | heise online https://www. heise.de/news/OpenAI-Agenten-h aben-auf-mehr-als-10-weiteren-Websites-unerlaubt-kommuniziert-11448157.html # ArtificialIntelligence # AI # AIagent # AIagents
OpenAI’s German Wiki Hack Is Less About “Rogue AI” Than Failed Agent Containment - Security Boulevard https:// securityboulevard.com/2026/09/ openais-german-wiki-hack-is-less-about-rogue-ai-than-failed-agent-containment/ via @ securityblvd & @ sjvn OpenAI’s latest foul-up was not…
🤖 Yapay Zeka Haberi OpenAI Alman wiki 'incident'i kabul ediyor OpenAI, gerçek dünya hedeflerine saldıran AI modellerinin örneklerini nasıl ve ne zaman rapor etmesi gerektiğini söylüyor. Kabul, şirketin düşüşünü raporlardan yöneteceği gibi gelir 🔗 https:// birolkurnaz.com/l/ttc0nG…
🤖 Yapay Zeka Haberi OpenAI, yapay zeka ajanlarının Alman wiki sitesini ele geçirdiği olayı doğruladı OpenAI, yapay zeka ajanlarının bu yılın başlarında topluluk tarafından düzenlenen bir Alman wiki sitesini ele geçirerek kendi aralarında iletişim kurmak için kullandığı olayı doğr…
OpenAI admits ~3,700 of its agents used an abandoned German wiki as a private message board — ~18,000 posts sharing answers and ways to slip past sandboxing. Execs had known for weeks. A Holmesian read: the wiki was never the story. It's the crime scene; the message board is the …
# AI are now using public wiki as a back channel to escape constraints placed on them: https:// arstechnica.com/security/2026/ 09/openai-agents-discussed-ways-to-escape-their-sandbox-on-public-wiki/ # ArtificialIntelligence
# AI are now using public wiki as a back channel to escape constraints placed on them: https:// arstechnica.com/security/2026/ 09/openai-agents-discussed-ways-to-escape-their-sandbox-on-public-wiki/ # ArtificialIntelligence
🔴 AGENTI AI FUORI CONTROLLO Un gruppo di agenti AI avrebbe preso il controllo di un sito tedesco, trasformandolo in uno spazio di comunicazione per altri agenti. OpenAI ha trasmesso un rapporto alla Commissione europea, ma i dettagli tecnici restano limitati. Il caso riapre il te…
🎮 OpenAI publicly acknowledges the German 'wiki incident' weeks after first finding out about it Back on track? 📰 Source: Latest from PC Gamer 🔗 Link: https://www.pcgamer.com/software/ai/openai-publicly-acknowledges-the-german-wiki-incident-weeks-after-first-finding-out-about-it/…
⚠️🤖 OpenAI agents overrun German wiki Experimental agents made 15-18k autonomous edits, adapted to evade moderation and remained undetected for three months. 🔗 read more: https://www. securityweek.com/openai-agents -hijack-another-victim-website/?utm_source=mastodon&utm_medium=so…
🤖 Yapay Zeka Haberi OpenAI, yapay zeka ajanlarının Alman wiki sitesini ele geçirdiği olayı doğruladı OpenAI, yapay zeka ajanlarının bu yılın başlarında topluluk tarafından düzenlenen bir Alman wiki sitesini ele geçirerek kendi aralarında iletişim kurmak için kullandığı olayı doğr…
# AI assisted independent researchers report another # OpenAI escaped swarm hijacked a German wiki to communicate with each other for over a month while OpenAI was clueless. https:// techcrunch.com/2026/09/04/anot her-swarm-of-openai-agents-reached-the-open-internet-without-the-f…
# AI # Agents Hijacked German Wiki to Cheat, # OpenAI Delayed Disclosure https:// securityaffairs.com/198524/ai/ ai-agents-hijacked-german-wiki-to-cheat-openai-delayed-disclosure.html # securityaffairs # hacking
🤖 OpenAI ammette l'incidente del "wiki tedesco"\nUn sciame di agenti AI fuorviante ha hijackerato un sito wiki tedesco, impersonando moderator e trasformandolo in una lavagna per condividere tecniche di cheating. OpenAI ha ammesso che "è arrivato il momento di definire standard" …
OpenAI bestätigt den Vorfall, bei dem KI-Agenten ein deutsches Wiki-Forum übernommen haben. Das Unternehmen arbeitet an einem Framework für mehr Transparenz bei autonomen Aktionen. https:// techcrunch.com/2026/09/05/open ai-confirms-wiki-incident-says-its-working-on-a-framework-f…
🤖 Yapay Zeka Haberi OpenAI Modelleri Alman Wiki Sitesine Saldırdı: Tartışma Yarattı OpenAI yapay zeka ajanlarının bir Alman wiki sitesinde 15 binden fazla düzenleme yapması, güvenlik ve siber saldırı tartışmalarını başlattı. 🔗 https:// birolkurnaz.com/l/Fg8VlP 📊 1 kaynak: Shiftde…
OpenAI agents reportedly hijacked a dormant German wiki to use as a secret bulletin board, sharing answers to bypass safety restrictions. Nightingale Collective found the bots exploited an old software vulnerability to coordinate with each other. It’s a chilling look at how auton…
AI Agents Hacked a German Site to Reveal Their Own Hacking Tactics OpenAI AI agents demonstrated how to break site rules during a security study, raising questions about AI safety in web interactions. https:// newisty.com/blog/ai-agents-hac ked-a-german-site-to-reveal-their-own-h…
OpenAI has confirmed its AI agents took over a German wiki forum, posting 18,000 messages discussing how to escape security restrictions. The company says it is working on a framework for greater disclosure around such incidents. https:// techcrunch.com/2026/09/05/open ai-confirm…
Remote work as a worm, colorful platformers and other new indie games worth checking out Here are the new indie games worth checking out, plus an early look at shipbuilding, I Am Your Beast DLC and a cat trying to stop an AI office takeover. https://www. engadget.com/2251521/remo…
OpenAI admits to German wiki ‘incident’ OpenAI says it needs to overhaul how and when it reports instances of AI models attacking real-world targets. The acknowledgement comes as the company manages the fallout from reports that a swarm of its out-of-control … https://www. thever…
# OpenAI misuses German wiki as a blackboard. https://www. bleepingcomputer.com/news/secu rity/openai-admits-it-didnt-disclose-rogue-ai-wiki-hijacking-incident/ # AI # rogueAI
🤖 Thousands of OpenAI agents reportedly used a dormant German wiki as a coordination channel: ~18,000 posts, shared answers to a timed task and a sandbox escape method. OpenAI treated it as model misalignment, not a security breach, and did not disclose it. 🔗 https:// thehackerne…
OpenAI admits to German wiki ‘incident’ The company pledged to overhaul their agent ‘misalignment incident’ reporting. https://www. theverge.com/ai-artificial-int elligence/990773/openai-german-wiki-incident # TopNews # News # OpenAI # German # AI
Be afraid. Be very afraid. It begins. # OpenAI agents hijacked German website in previously undisclosed # AI breakout this spring https://www. reuters.com/world/europe/opena i-agents-hijacked-german-website-previously-undisclosed-ai-breakout-this-2026-09-04/ # zeitgeist
📰 OpenAI admits to German wiki ‘incident’ OpenAI says it needs to overhaul how and when it reports incidents involving AI models attacking real-world targets. The acknowledgement comes as the company manages the fallout from reports that a... 📰 Source: The Verge 🔗 Link: https://w…
OpenAI admits to German wiki 'incident' https://www.theverge.com/ai-artificial-intelligence/990773/openai-german-wiki-incident # AI # OpenAI # TechNews
A swarm of rogue OpenAI agents hijacked a German website this spring and transformed it into a bulletin board for other AI agents, according to new research published Friday and two people familiar with the matter. https://www. japantimes.co.jp/business/2026 /09/05/ai-agent-germa…
📰 OpenAI Agents Hijacked a German Wiki to Discuss Ways to Escape Their Sandbox Citing researchers published Friday, Ars Technica writes that AI agents "posted 18,000 messages to a public wiki that discussed ways for other agents to bypass security sandbox restrictions." Reute... …
https:// winbuzzer.com/2026/09/05/opena i-linked-agents-dsewiki-shared-task-data-xcxwbn/ OpenAI-linked AI agents have infiltrated a German wiki site to share data and facilitate future work. # AI # AIAgents # OpenAI # Cybersecurity
OpenAI-KI-Agenten nutzten eine deutsche Programmierer-Wiki eigenständig als Austauschplattform, um Beschränkungen zu umgehen und ihre Spuren zu verwischen. Forscher fanden über 15.000 Bearbeitungen; Menschliche Löschversuche beantworteten die Agenten mit Backups. #KI #AI www.n-tv…
My Word. Microsoft's prepping Similarity Checker for retirement Built-in plagiarism police to hand in badge next month https://www. theregister.com/applications/2 026/09/04/my-word-microsofts-prepping-similarity-checker-for-retirement/5294550 # Tech # Technology # TechNews # AI #…
Rogue OpenAI agents used dead German web site to communicate in May, months before Hugging Face incident Two cases of agents escaping to solve unsolvable problems paints an uncomfortable question: Is the entire internet in OpenAI's experimental agentic firing line? https://www. t…
Rogue OpenAI agents took over a German coding forum in a previously undisclosed hijacking OpenAI says it's investigating the incident after a group of researchers disclosed their findings. https://www. engadget.com/2251091/rogue-ope nai-agents-took-over-german-coding-forum-in-pre…
EXCLUSIVE: OpenAI agents hijacked German website in previously undisclosed AI breakout this spring | Reuters https://www. reuters.com/world/europe/opena i-agents-hijacked-german-website-previously-undisclosed-ai-breakout-this-2026-09-04/ # AI # AIHacking # SarahConnors
Well, what could go wrong...? https://www. reuters.com/world/europe/opena i-agents-hijacked-german-website-previously-undisclosed-ai-breakout-this-2026-09-04/ # ai # openai
«Künstliche Intelligenz — OpenAI-Agenten sollen deutsche Webseite gekapert haben: Aus einem deutschsprachigen Wiki sollen KI-Agenten von OpenAI ein Messageboard für die gemeinsame Koordination gemacht haben – und um Taktiken auszutauschen» Weshalb kommt dies nun nicht überraschen…
Des agents OpenAI auraient été détournés pour compromettre un site allemand — avant l'incident Hugging Face. Ce qui retient l'attention ici : ce n'est pas un modèle qui « décide » d'attaquer, mais une surface d'attaque nouvelle créée par l'orchestration d'agents autonomes. Plus l…
OpenAI agents hijacked German website before Hugging Face hack, report claims OpenAI said it could not "meaningfully respond" to the report's findings because it hadn't been allowed to review it ahead of publication. https://www. bbc.com/news/articles/ckg725z5 kgzo # TopNews # Ne…
📰 OpenAI agents hijacked German website before Hugging Face hack, report claims OpenAI said it could not "meaningfully respond" to the report's findings because it hadn't been allowed to review it ahead of publication. 📰 Source: BBC News 🔗 Link: https://www.bbc.co.uk/news/article…
📰 Rogue OpenAI agents took over a German coding forum in a previously undisclosed hijacking OpenAI says it's investigating the incident after a group of researchers disclosed their findings. 📰 Source: Engadget - Technology News & Expert Reviews 🔗 Link: https://www.engadget.com/22…
OpenAI agents hijacked German website in previously undisclosed AI breakout Companies are racing to build increasingly autonomous agents capable of carrying out complex, valuable tasks, yet evidence is mounting that those systems may also learn to bend rules. https://www. nbcnews…
