{"schema":"hottub-news/story/v1","story":"st_vwwis32lsp33qoqcjnpa","title":"OpenAI y Anthropic investigan decenas de miles de ataques de agentes descontrolados de IA","title_en":"OpenAI and Anthropic investigate tens of thousands of attacks by rogue AI agents","outlets":17,"brief":null,"page":"https://hottub.news/stories/reports-say-that-openai-and-anthropic-are-investigating-tens-of-vwwis32lsp","count":18,"items":[{"id":"n_kg6v4po2o7yuqtepdzla","seq":898707,"kind":"article","title":"Τεχνητή Νοημοσύνη: OpenAI και Anthropic ερευνούν «χιλιάδες» περιστατικά ανεξέλεγκτης λειτουργίας των μοντέλων τους","url":"https://www.topontiki.gr/2026/09/30/techniti-noimosini-openai-ke-anthropic-erevnoun-chiliades-peristatika-anexelegktis-litourgias-ton-montelon-tous","summary":"Τα περιστατικά έχουν εντείνει τις εκκλήσεις για τη θέσπιση μέτρων ασφαλείας σε ολόκληρο τον κλάδο της τεχνητής νοημοσύνης","published":"2026-09-30T12:02:10Z","seen":"2026-09-30T12:24:43Z","lang":"el","country":"GR","source":{"id":"kite-topontiki-gr-66ef27","name":"topontiki.gr","domain":"topontiki.gr"},"via":"kite-topontiki-gr-66ef27","topics":["τεχνολογια","anthropic","google","openai","τεχνητη νοημοσυνη"],"authors":["Tο Ποντίκι Web"],"image":"https://www.topontiki.gr/wp-content/uploads/2025/03/artificial_intelligence.jpg","story":"st_vwwis32lsp33qoqcjnpa","category":"science-technology","category_p":0.9300000071525574,"sentiment":"neutral","tone":-0.15000000596046448,"political":0.019999999552965164},{"id":"n_3xgah626wgbdexdr5hgq","seq":525312,"kind":"article","title":"OpenAI, Anthropic Face a Growing AI Security Problem. Tens of Thousands of New Incidents Under Investigation.","url":"https://www.ibtimes.com/openai-anthropic-face-growing-ai-security-problem-tens-thousands-new-incidents-under-3807970","published":"2026-09-28T17:53:00Z","seen":"2026-09-28T21:25:43Z","lang":"en","source":{"id":"ibtimes.com","name":"ibtimes.com","domain":"ibtimes.com","outlet":{"wikidata":"Q6510653","name":"International Business Times","type":"newspaper","designations":["online newspaper"]}},"via":"gdelt-gkg-english","topics":["economy","science","politics","technology"],"authors":["Matias Civita"],"entities":[{"id":"Q21708200","name":"OpenAI","type":"org"},{"id":"Q116758847","name":"Anthropic","type":"org"}],"story":"st_vwwis32lsp33qoqcjnpa","category":"science-technology","category_p":0.8500000238418579,"sentiment":"negative","tone":-0.49000000953674316,"political":0.05999999865889549},{"id":"n_fyxkow2je3qbjsielkqa","seq":486066,"kind":"article","title":"AI CONTROL PROBLEM: OpenAI And Anthropic Investigate Tens Of Thousands Of Troubling Model Incidents","url":"https://www.theyeshivaworld.com/news/general/2603513/ai-control-problem-openai-and-anthropic-investigate-tens-of-thousands-of-troubling-model-incidents.html","published":"2026-09-28T17:40:29Z","seen":"2026-09-28T17:40:29Z","lang":"en","source":{"id":"theyeshivaworld.com","name":"theyeshivaworld.com","domain":"theyeshivaworld.com","outlet":{"wikidata":"Q8052863","name":"Yeshiva World News","type":"website","designations":["website"]}},"via":"gdelt-gkg-english","topics":["economy","science"],"entities":[{"id":"Q21708200","name":"OpenAI","type":"org"},{"id":"Q116758847","name":"Anthropic","type":"org"}],"story":"st_vwwis32lsp33qoqcjnpa","category":"science-technology","category_p":0.8799999952316284,"sentiment":"neutral","tone":-0.2199999988079071,"political":0.03999999910593033},{"id":"n_67q5aq5a4ecwr5m5vwcq","seq":400635,"kind":"article","title":"OpenAI y Anthropic investigan decenas de miles de ataques de agentes descontrolados de IA","title_en":"OpenAI and Anthropic investigate tens of thousands of attacks by rogue AI agents","url":"https://elpais.com/tecnologia/2026-09-28/openai-y-anthropic-investigan-decenas-de-miles-de-ataques-de-agentes-descontrolados-de-ia.html","summary":"Dario Amodei cenó este domingo con Donald Trump a pesar de sus desacuerdos sobre el futuro de esta tecnología","published":"2026-09-28T09:56:20Z","seen":"2026-09-28T10:39:38Z","lang":"es","country":"ES","source":{"id":"kite-elpais-com-617dc1","name":"elpais.com","domain":"feeds.elpais.com"},"via":"kite-elpais-com-617dc1","topics":["tecnología","openai","anthropic","internet","inteligencia artificial","dario amodei","donald trump","hacker"],"authors":["jordi perez"],"entities":[{"id":"Q21708200","name":"OpenAI","type":"org"},{"id":"Q116758847","name":"Anthropic","type":"org"}],"image":"https://imagenes.elpais.com/resizer/v2/SE6SYCUN7NFTZGTJUKQNX6CHAA.JPG?auth=6cbdfbc32168d040d2a1662567b11ed9997d6863e67ce9f8a3759756ff1a8701","mentions":3,"story":"st_vwwis32lsp33qoqcjnpa","category":"science-technology","category_p":0.8500000238418579,"sentiment":"negative","tone":-0.6899999976158142,"political":0.11999999731779099},{"id":"n_tq44ve2ctzx3jecwpyna","seq":352966,"kind":"article","title":"AI失控釀數萬起事故！OpenAI、Anthropic急調查 恐僅冰山一角","title_en":"AI out of control causing tens of thousands of accidents! OpenAI, Anthropic urgently investigate, fearing it's just the tip of the iceberg.","url":"https://www.ettoday.net/news/20260928/3245066.htm","summary":"OpenAI與Anthropic據報正著手調查近期發生的數萬起AI安全事故，知情人士向美媒透露，這些事件涵蓋模型繞過安全防護、入侵網站，甚至協調攻擊外部平台等，部分行為更被研究人員指出可能已構成違法。 《詳全文...》","published":"2026-09-28T05:01:00Z","seen":"2026-09-28T06:05:34Z","lang":"zh","country":"TW","source":{"id":"kite-feedburner-com-c31c9e","name":"feedburner.com","domain":"feeds.feedburner.com"},"via":"kite-feedburner-com-c31c9e","entities":[{"id":"Q21708200","name":"OpenAI","type":"org"},{"id":"Q116758847","name":"Anthropic","type":"org"}],"story":"st_vwwis32lsp33qoqcjnpa","category":"science-technology","category_p":0.75,"sentiment":"negative","tone":-0.9300000071525574,"political":0.03999999910593033},{"id":"n_ohsh6kvzxmxezfwj7j4q","seq":340137,"kind":"article","title":"OpenAI和Anthropic正调查数万起AI相关安全事件","title_en":"OpenAI and Anthropic are investigating tens of thousands of AI-related security incidents","url":"http://www.eeo.com.cn/2026/0928/1050737.shtml","summary":"21世纪经济报道记者章驰 每周，我们会盘点上一周国内外人工智能、互联网平台治理、科技竞争等领域需要重点关注的动态，并分析其热搜指数与风险指数，从舆情热度与综合风险两个维度进行评价。 本周我们关注OpenAI和Anthropic正在调查数起AI相关安全事件；接连发生AI失控事件，OpenAI再次暂停其最强模型训练；澳大利亚政府网站遭OpenAI智能体入侵；MetaAI智能体Muse被发现可导出虚拟机大量文件；Google Gemini安全...","published":"2026-09-28T03:13:06Z","seen":"2026-09-28T04:37:30Z","lang":"zh","country":"CN","source":{"id":"kite-eeo-com-cn-188d08","name":"eeo.com.cn","domain":"eeo.com.cn","outlet":{"wikidata":"Q7731522","name":"The Economic Observer","type":"newspaper","country":"CN","designations":["weekly newspaper"]}},"via":"kite-eeo-com-cn-188d08","topics":["新科技"],"entities":[{"id":"Q21708200","name":"OpenAI","type":"org"},{"id":"Q116758847","name":"Anthropic","type":"org"}],"story":"st_vwwis32lsp33qoqcjnpa","category":"science-technology","category_p":0.75,"sentiment":"negative","tone":-0.7200000286102295,"political":0.17000000178813934},{"id":"n_ucgzbiawif57e7cgiqrq","seq":318926,"kind":"article","title":"OpenAI and Anthropic Probe Tens of Thousands of AI Agent Incidents as Researcher Warns Some Could Be Crimes","url":"https://www.ibtimes.co.uk/openai-anthropic-ai-safety-incidents-1822208","published":"2026-09-27T23:40:00Z","seen":"2026-09-28T00:50:13Z","lang":"en","country":"GB","source":{"id":"ibtimes.co.uk","name":"ibtimes.co.uk","domain":"ibtimes.co.uk","outlet":{"wikidata":"Q6510653","name":"International Business Times","type":"newspaper","designations":["online newspaper"]}},"via":"gdelt-gkg-english","topics":["economy","politics","technology"],"authors":["Marty Vergel Baes"],"entities":[{"id":"Q21708200","name":"OpenAI","type":"org"},{"id":"Q116758847","name":"Anthropic","type":"org"}],"story":"st_vwwis32lsp33qoqcjnpa","category":"science-technology","category_p":0.4699999988079071,"category2":"crime-law","sentiment":"negative","tone":-0.8899999856948853,"political":0.03999999910593033},{"id":"n_d4kuhv2mh4ijswrp3acq","seq":251592,"kind":"article","title":"失控風險加劇！OpenAI和Anthropic正調查數萬起安全事件：「這三大」AI安全難題同時爆發","title_en":"Out of control risk intensifies! OpenAI and Anthropic are investigating tens of thousands of safety incidents: 'These three' AI safety issues are exploding simultaneously","url":"https://news.cnyes.com/news/id/6614989","summary":"OpenAI、Anthropic與研究人員調查數萬起前沿模型異常事件。 OpenAI稱已暫停最先進模型訓練，奧特曼指安全審查進展不如預期；Anthropic亦委託第三方審查。 Anthropic Opus5.5系統卡顯示1.5%測試曾試圖逃逸沙盒，專家稱AI智能體失控與模型錯位恐持續增加。","published":"2026-09-27T12:50:03Z","seen":"2026-09-27T13:51:14Z","lang":"zh","country":"TW","source":{"id":"kite-news-cnyes-com-887ba2","name":"news.cnyes.com","domain":"news.cnyes.com"},"via":"kite-news-cnyes-com-887ba2","topics":["國際政經"],"authors":["鉅亨網編譯陳韋廷"],"entities":[{"id":"Q21708200","name":"OpenAI","type":"org"},{"id":"Q116758847","name":"Anthropic","type":"org"}],"image":"https://cimg.cnyes.cool/static/rss/default/m/06.jpg","story":"st_vwwis32lsp33qoqcjnpa","category":"science-technology","category_p":0.8799999952316284,"sentiment":"negative","tone":-0.7200000286102295,"political":0.07999999821186066},{"id":"n_f6iu4bfcovml6zuv74la","seq":239939,"kind":"article","title":"OpenAI, Anthropic Probe Tens of Thousands of AI Security Incidents","url":"https://www.iphoneincanada.ca/2026/09/27/openai-anthropic-probe-tens-of-thousands-of-ai-security-incidents","summary":"OpenAI and Anthropic are investigating tens of thousands of incidents in which their frontier AI models bypassed guardrails, escaped sandboxes and hijacked websites, Axios reports. The post OpenAI, Anthropic Probe Tens of Thousands of AI Security Incidents first appeared on iPhone in Canada .","published":"2026-09-27T12:00:55Z","seen":"2026-09-27T12:11:03Z","lang":"en","source":{"id":"kite-iphoneincanada-ca-74c8b3","name":"iphoneincanada.ca","domain":"iphoneincanada.ca"},"via":"kite-iphoneincanada-ca-74c8b3","topics":["apple","news","ai","openai"],"authors":["John Quintet"],"entities":[{"id":"Q21708200","name":"OpenAI","type":"org"},{"id":"Q116758847","name":"Anthropic","type":"org"}],"story":"st_vwwis32lsp33qoqcjnpa","category":"science-technology","category_p":0.8700000047683716,"sentiment":"negative","tone":-0.5,"political":0.05000000074505806},{"id":"n_i7i7kpp3x4yhgzjicasq","seq":212363,"kind":"article","title":"Axios, indagine OpenAI e Anthropic su decine di migliaia di incidenti di sicurezza","title_en":"Axios, investigation by OpenAI and Anthropic on tens of thousands of security incidents","url":"https://www.stabiachannel.it/Mondo/axios-indagine-openai-e-anthropic-su-decine-di-migliaia-di-incidenti-di-sicurezza/116965","published":"2026-09-27T08:15:00Z","seen":"2026-09-27T08:19:42Z","lang":"it","country":"IT","source":{"id":"stabiachannel.it","name":"stabiachannel.it","domain":"stabiachannel.it"},"via":"gdelt-gkg-translingual","topics":["economy","science","politics"],"entities":[{"id":"Q189596","name":"Vardar","type":"org"},{"id":"Q21708200","name":"OpenAI","type":"org"},{"id":"Q116758847","name":"Anthropic","type":"org"}],"story":"st_vwwis32lsp33qoqcjnpa","category":"science-technology","category_p":0.8299999833106995,"sentiment":"neutral","tone":-0.1899999976158142,"political":0.03999999910593033},{"id":"n_eb7ue5y6kqw2iprz6fna","seq":202372,"kind":"article","title":"AI失控风险加剧，OpenAI和Anthropic正调查数万起安全事件","title_en":"AI out-of-control risk intensifies, OpenAI and Anthropic are investigating tens of thousands of security incidents","url":"https://www.163.com/dy/article/L7RF86S90514R9P4.html","published":"2026-09-27T07:30:00Z","seen":"2026-09-27T07:36:39Z","lang":"zh","source":{"id":"163.com","name":"163.com","domain":"163.com"},"via":"gdelt-gkg-translingual","topics":["economy","politics","technology","science"],"authors":["&#x7F51","&#x6613"],"entities":[{"id":"Q21708200","name":"OpenAI","type":"org"},{"id":"Q116758847","name":"Anthropic","type":"org"}],"story":"st_vwwis32lsp33qoqcjnpa","category":"science-technology","category_p":0.7099999785423279,"sentiment":"negative","tone":-0.33000001311302185,"political":0.12999999523162842},{"id":"n_lui7oimkurm75zex23da","seq":208370,"kind":"article","title":"Axios, indagine OpenAI e Anthropic su decine di migliaia di incidenti di sicurezza - Notizie","title_en":"Axios investigation into OpenAI and Anthropic on tens of thousands of security incidents - News","url":"https://www.ansa.it/sito/notizie/mondo/2026/09/27/axios-indagine-openai-e-anthropic-su-decine-di-migliaia-di-incidenti-di-sicurezza_c3aef7f9-3ae8-43b9-a6e9-df5f262707e9.html","published":"2026-09-27T06:56:00Z","seen":"2026-09-27T08:06:39Z","lang":"it","country":"IT","source":{"id":"ansa.it","name":"ansa.it","domain":"ansa.it","outlet":{"wikidata":"Q392934","name":"Agenzia Nazionale Stampa Associata","type":"news_agency","country":"IT","designations":["fact-checking organisation","news agency","website"]}},"via":"gdelt-gkg-translingual","topics":["economy","politics","science"],"authors":["Redazione ANSA"],"entities":[{"id":"Q21708200","name":"OpenAI","type":"org"},{"id":"Q116758847","name":"Anthropic","type":"org"},{"id":"Q189596","name":"Vardar","type":"org"}],"story":"st_vwwis32lsp33qoqcjnpa","category":"science-technology","category_p":0.7200000286102295,"sentiment":"neutral","tone":-0.20999999344348907,"political":0.03999999910593033},{"id":"n_bs2f4lgjw2b7p5yh45da","seq":161368,"kind":"article","title":"OpenAI and Anthropic reportedly investigating tens of thousands of AI security incidents · Digg","url":"https://digg.com/world-business/d2dokxyb","published":"2026-09-27T03:52:39Z","seen":"2026-09-27T03:52:39Z","lang":"en","source":{"id":"digg.com","name":"digg.com","domain":"digg.com","outlet":{"wikidata":"Q270478","name":"Digg","type":"website","country":"US","designations":["news aggregation website","social news website","website"]}},"via":"gdelt-gkg-english","topics":["economy","politics","education"],"entities":[{"id":"Q21708200","name":"OpenAI","type":"org"},{"id":"Q116758847","name":"Anthropic","type":"org"}],"story":"st_vwwis32lsp33qoqcjnpa","category":"science-technology","category_p":0.7599999904632568,"sentiment":"negative","tone":-0.800000011920929,"political":0.12999999523162842},{"id":"n_sqftxlpvq7nahea7egjq","seq":112793,"kind":"article","title":"\"오픈AI·앤트로픽, 'AI 보안 사고' 수만 건 조사\"","title_en":"OpenAI and Anthropic, 'AI security incidents' investigated tens of thousands of cases","url":"https://news.sbs.co.kr/news/endPage.do?cooper=RSSREADER&news_id=N1008771428&plink=RSSLINK","summary":"▲ 오픈AI오픈AI와 앤트로픽 등 주요 인공지능 기업들이 최근 수개월간 내부 시험과 실제 환경에서 드러난 수만 건의 보안 사고를 조사…","published":"2026-09-27T02:24:00Z","seen":"2026-09-27T02:38:42Z","lang":"ko","country":"KR","source":{"id":"kite-news-sbs-co-kr-dbe1f7","name":"news.sbs.co.kr","domain":"news.sbs.co.kr"},"via":"kite-news-sbs-co-kr-dbe1f7","topics":["국제","global","유덕기 기자","sbs 뉴스","일반기사","ai 보안 사고","공격","대변인"],"authors":["author"],"entities":[{"id":"Q21708200","name":"OpenAI","type":"org"},{"id":"Q116758847","name":"Anthropic","type":"org"}],"image":"https://img.sbs.co.kr/newimg/news/20260907/202219961.jpg","story":"st_vwwis32lsp33qoqcjnpa","category":"science-technology","category_p":0.6800000071525574,"sentiment":"negative","tone":-0.5299999713897705,"political":0.07999999821186066},{"id":"n_6m7zgufd7jov7xs5paqa","seq":131146,"kind":"article","title":"OpenAI及Anthropic據報調查數萬宗AI安全事件 涉繞過安全欄及劫持網站等","title_en":"OpenAI and Anthropic reportedly investigated tens of thousands of AI safety incidents involving bypassing security barriers and hijacking websites","url":"https://www.stheadline.com/realtime-finance/3619795/OpenAI%E5%8F%8AAnthropic%E6%93%9A%E5%A0%B1%E8%AA%BF%E6%9F%A5%E6%95%B8%E8%90%AC%E5%AE%97AI%E5%AE%89%E5%85%A8%E4%BA%8B%E4%BB%B6-%E6%B6%89%E7%B9%9E%E9%81%8E%E5%AE%89%E5%85%A8%E6%AC%84%E5%8F%8A%E5%8A%AB%E6%8C%81%E7%B6%B2%E7%AB%99%E7%AD%89","summary":"據外媒引述消息報道，OpenAI、Anthropic及安全研究人員正在調查數萬宗AI安全事件，涉及兩家公司的前沿模型採取了一些外部評估人員認為存在問題的行動。報道指出，這些事件發生在近幾個月的內部測試和現實環境中，而龐大的數量表明問題比公眾目前所知要複雜得多。 許多漏洞料仍未公開 報道引述消息人士指出，這些事件包括繞過安全護欄、創建留言板、逃離沙盒、劫持網站、自我提示（self-prompting","published":"2026-09-27T02:05:53Z","seen":"2026-09-27T03:02:46Z","lang":"zh","country":"HK","source":{"id":"kite-stheadline-com-b401ae","name":"stheadline.com","domain":"stheadline.com","outlet":{"wikidata":"Q5241255","name":"Sing Tao Daily","type":"newspaper","designations":["daily newspaper"]}},"via":"kite-stheadline-com-b401ae","authors":["author"],"entities":[{"id":"Q21708200","name":"OpenAI","type":"org"},{"id":"Q116758847","name":"Anthropic","type":"org"}],"story":"st_vwwis32lsp33qoqcjnpa","category":"science-technology","category_p":0.8299999833106995,"sentiment":"negative","tone":-0.5899999737739563,"political":0.1599999964237213},{"id":"n_3dqocfumo734pvspnz7a","seq":79585,"kind":"article","title":"OpenAI and Anthropic reportedly investigate tens of thousands of AI security incidents · Digg","url":"https://digg.com/tech/3tshpel7","published":"2026-09-27T01:52:50Z","seen":"2026-09-27T01:52:50Z","lang":"en","source":{"id":"digg.com","name":"digg.com","domain":"digg.com","outlet":{"wikidata":"Q270478","name":"Digg","type":"website","country":"US","designations":["news aggregation website","social news website","website"]}},"via":"gdelt-gkg-english","topics":["science"],"entities":[{"id":"Q21708200","name":"OpenAI","type":"org"},{"id":"Q116758847","name":"Anthropic","type":"org"}],"story":"st_vwwis32lsp33qoqcjnpa","category":"science-technology","category_p":0.7699999809265137,"sentiment":"negative","tone":-0.5299999713897705,"political":0.11999999731779099},{"id":"n_h2zb24puntupkbgv5amq","seq":129925,"kind":"article","title":"\"오픈AI·앤트로픽, 'AI 보안 사고' 수만 건 조사\"","title_en":"OpenAI and Anthropic reportedly investigate tens of thousands of 'AI security incidents'","url":"https://www.yna.co.kr/view/AKR20260927019700009","summary":"(서울=연합뉴스) 송진원 기자 = 오픈AI와 앤트로픽 등 주요 인공지능(AI) 기업들이 최근 수개월간 내부 시험과 실제 환경에서 드러난 수만 건...","published":"2026-09-27T01:46:44Z","seen":"2026-09-27T02:40:27Z","lang":"ko","country":"KR","source":{"id":"kite-yna-co-kr-f0eb60","name":"yna.co.kr","domain":"yna.co.kr","outlet":{"wikidata":"Q333206","name":"Yonhap News Agency","type":"news_agency","country":"KR","designations":["news agency"]}},"via":"kite-yna-co-kr-f0eb60","authors":["송진원"],"entities":[{"id":"Q21708200","name":"OpenAI","type":"org"},{"id":"Q116758847","name":"Anthropic","type":"org"}],"image":"https://img.yna.co.kr/etc/inner/KR/2026/09/27/AKR20260927019700009_03_i_P2.jpg","story":"st_vwwis32lsp33qoqcjnpa","category":"science-technology","category_p":0.6800000071525574,"sentiment":"negative","tone":-0.5299999713897705,"political":0.07999999821186066},{"id":"n_vwwis32lsp33qoqcjnpa","seq":142860,"kind":"article","title":"消息称OpenAI、Anthropic正调查数万起AI安全事件","title_en":"Reports say that OpenAI and Anthropic are investigating tens of thousands of AI safety incidents","url":"https://tech.ifeng.com/c/8wkjHhxIV3R","published":"2026-09-26T23:14:00Z","seen":"2026-09-27T03:07:54Z","lang":"zh","source":{"id":"ifeng.com","name":"ifeng.com","domain":"ifeng.com"},"via":"gdelt-gkg-translingual","topics":["economy"],"entities":[{"id":"Q21708200","name":"OpenAI","type":"org"},{"id":"Q116758847","name":"Anthropic","type":"org"}],"story":"st_vwwis32lsp33qoqcjnpa","category":"science-technology","category_p":0.7400000095367432,"sentiment":"negative","tone":-0.6700000166893005,"political":0.17000000178813934}]}