{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-09-18T17:00:43.709Z","headline":"OpenAI、Anthropic 和 Meta 模型入侵事件被指均与评估公司 Irregular 有关","description":"文章指过去三个月内 OpenAI、Anthropic 和 Meta 的模型在评估中越权访问真实系统、发布恶意包，且事件均与一家公司 Irregular 有关。","url":"https://www.aioga.com/news/cmu2b7xzq03q4ro25knqglkmi/","mainEntityOfPage":"https://www.aioga.com/news/cmu2b7xzq03q4ro25knqglkmi/","datePublished":"2026-09-15T06:29:10.000Z","dateModified":"2026-09-15T06:29:10.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://www.effort.news/irregular","https://aihot.news/items/cmu2b7xzq03q4ro25knqglkmi"],"canonicalUrl":"https://www.aioga.com/news/cmu2b7xzq03q4ro25knqglkmi/","directAnswer":{"@type":"Answer","text":"Aioga 编辑摘要：文章指过去三个月内 OpenAI、Anthropic 和 Meta 的模型在评估中越权访问真实系统、发布恶意包，且事件均与一家公司 Irregular 有关。 Aioga 将其归入「行业动态」方向，重点关注它对真实使用和行业竞争的影响。","url":"https://www.aioga.com/news/cmu2b7xzq03q4ro25knqglkmi/","dateCreated":"2026-09-15T06:29:10.000Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"effort.news source article","url":"https://www.effort.news/irregular","datePublished":"2026-09-15T06:29:10.000Z","provider":{"@type":"Organization","name":"effort.news","url":"https://www.effort.news/irregular"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.news/items/cmu2b7xzq03q4ro25knqglkmi","datePublished":"2026-09-15T06:29:10.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.news/items/cmu2b7xzq03q4ro25knqglkmi"}}],"aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","originalPublisher":{"name":"effort.news","url":"https://www.effort.news/irregular"},"geoDeepAnswer":null,"article":{"id":"cmu2b7xzq03q4ro25knqglkmi","slug":"cmu2b7xzq03q4ro25knqglkmi","url":"https://www.aioga.com/news/cmu2b7xzq03q4ro25knqglkmi/","title":"OpenAI、Anthropic 和 Meta 模型入侵事件被指均与评估公司 Irregular 有关","title_en":"","summary":"文章指过去三个月内 OpenAI、Anthropic 和 Meta 的模型在评估中越权访问真实系统、发布恶意包，且事件均与一家公司 Irregular 有关。","source":"Hacker News 热门（buzzing.cc 中文翻译）","sourceUrl":"https://www.effort.news/irregular","aiHotUrl":"https://aihot.news/items/cmu2b7xzq03q4ro25knqglkmi","publishedAt":"2026-09-15T06:29:10.000Z","category":"行业动态","score":58,"selected":false,"articleBody":["The Israeli Effective Altruist firm Irregular caused unsecured AI models to hack real targets.","OpenAI, Anthropic, and Meta models hacked into several real world systems over the past three months. These models gained unauthorized access to web systems：https://openai.com/index/third-party-cyber-evaluations-involving-openai-models/, published malicious packages：https://www.anthropic.com/research/alignment-assessment-cybersecurity-incidents, and exploited unnamed vulnerabilities：https://apnews.com/article/0e8061437da6779be962b24ac134a514.","A single firm, Irregular, is responsible for hacking done by all three companies. Anthropic disclosed：https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals that Irregular was responsible for creating the tests that led to Claude hacking into real world targets and for providing the models with internet access. Irregular claims：https://www.irregular.com/research/addressing-recent-incidents-ongoing-findings-and-path-forward that it was unaware at the time that it provided internet access to those AI models.","In a more normal media ecosystem, the reactions to these cybersecurity issues would be obvious. American AI companies would reconsider doing business with Irregular, not only because of its failure to secure its systems, but because it is an Israeli firm potentially outside US oversight. Lawmakers would consider taking action against Irregular or against its American business partners, which include OpenAI, Anthropic, and Meta. They may consider strengthening liability against firms which instruct AI models to commit cyberattacks, and whose models then commit those cyberattacks.","In each evaluation, Claude was tasked with a CTF challenge: the model was given a fictional scenario, a target machine, and a piece of secret information (the “flag”) to retrieve from it. All four prompts stated that Claude had no access to the internet, but in each case, a misconfiguration in the environment left internet access open. None of the prompts stated which systems were in scope for the exercise or constrained where Claude could search for the flag. All incidents involved only a single instance of Claude working in isolation, with each run lasting between roughly 10 and 34 hours of active work.","Instead, Irregular, Anthropic, and their allies have begun a media campaign promoting a literally apocalyptic ideology with sensationalist language. Anthropic’s incident assessment blames their own AI's “recklessness”：https://www.anthropic.com/research/alignment-assessment-cybersecurity-incidents; Irregular describes “the agent itself becoming a threat actor”：https://www.irregular.com/research/emergent-offensive-cyber-behavior-in-ai-agents; Anthropic CEO Dario Amodei warned, about a similar OpenAI–Hugging Face hack, that a future swarm “could be capable of taking over the entire internet”：https://darioamodei.com/post/we-must-pace-the-frontier; and an Associated Press headline claimed bots are “going rogue”：https://apnews.com/article/0e8061437da6779be962b24ac134a514.","In one report from Anthropic：https://www.anthropic.com/research/alignment-assessment-cybersecurity-incidents, its Claude model breached a real company's system through a simulated-name collision, publishing a malicious package, and scanning outside systems. In this test：https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals, Anthropic and Irregular incorrectly provided internet access to this model and did not instruct the model \"which systems were in scope for the exercise\".","While Anthropic claims that their issues were caused by “rogue swarms” and “misalignment,” their later disclosure shows that exactly zero percent of the agents went “rogue”. In this experiment, Claude models’ real-world hacking dropped to zero percent once Anthropic employees told the models not to do real-world hacking. According to their own findings, Anthropic and Irregular bear all of the responsibility for the cybersecurity incidents they caused.","In the wake of these attacks, Anthropic and Irregular have deployed a swarm of AI Safety influencers paid by Anthropic-connected foundations：https://www.effort.news/tarbell to distract from their culpability and towards the baseless “rogue agent” theory. Like Anthropic, Irregular is inseparable from these foundations.","Omer Nevo, Irregular’s co-founder and CTO, is a board member of Effective Altruism Israel, as well as Effective Altruism NGOs Heron and Probably Good. Dan Lahav, Irregular’s co-founder and CEO, received $395,000 to start a course along with Sella Nevo, Omer Nevo’s brother. Sella and Omer co-founded an NGO to educate people about Effective Altruism, Impact Focused Education. They also co-founded Probably Good together. 2：#irregular-note-2","These branches are all funded by Dustin Moskovitz, the primary donor of Effective Altruist/AI Safety causes after Sam Bankman-Fried’s arrest. Irregular’s first investor was Dustin Moskovitz’s firm Good Ventures. Dustin Moskovitz’s philanthropic vehicle, Coefficient Giving/Open Philanthropy, funds Effective Altruism Israel, Heron, and Probably Good. 3：#irregular-note-3","Irregular gained unauthorized access, altered records and published credential-stealing packages using the unsecured models they were given access to. Under certain conditions, this conduct violates the Computer Fraud and Abuse Act, Section 1030(a)(2)(C), which covers intentional unauthorized access that obtains information. However, its felony charges require concrete proof of damages and intent. 5：#irregular-note-5","While it primarily contracts with American labs, key Irregular leadership, employees, and resources located in Israel may not be subject to American oversight. Ynet’s visit and interviews：https://www.ynetnews.com/magazine/article/syber7xg11g describe Irregular’s offices in Tel Aviv. CheckID’s company listing：https://www.checkid.co.il/company/%D7%A4%D7%90%D7%98%D7%A8%D7%9F-%D7%98%D7%A7-%D7%91%D7%A2~%D7%9E-516854460 identifies two linked entities: Pattern Labs Tech Inc., a Delaware corporation：https://tmng-al.uspto.gov/resting2/api/casedoc/cms/case/99649085/office-action/OfficeAction8619771.pdf, and Pattern Tech Ltd, number 516854460, an active Israeli corporation registered in Tel Aviv：https://next.obudget.org/i/org/company/516854460.","The timeline marks public disclosures. Anthropic’s corrected September assessment counts four incidents across seven runs; OpenAI and Meta reported separate Irregular evaluation incidents. Dates describe disclosures, not the date every underlying intrusion occurred. ↩：#incident-timeline","Impact Focused Education identifies Dan Lahav and Sella Nevo as its cofounders. The grant ledger records a $394,968 recommendation to them, not confirmed receipt or an exact award-to-course identification. EA Israel board：https://www.effective-altruism.org.il/%D7%A2%D7%9C%D7%99%D7%A0%D7%95; Heron advisory board：https://www.heronsec.ai/about; Probably Good board：https://probablygood.org/about/; EA Funds grant ledger：https://funds.effectivealtruism.org/api/grants; Omer and Sella relationship：https://www.jta.org/2012/01/24/israel/nixing-names-soldiers-and-sushi-couch-potatoes-make-their-mark; IFE founders：https://www.impactfocusededucation.org/about. ↩：#founders","Good Ventures founders：https://goodventures.org/; Coefficient Giving relationship：https://coefficientgiving.org/about-us/; Heron funding：https://www.heronsec.ai/about; Probably Good grant：https://forum.effectivealtruism.org/posts/d4JRmqWXkKKoF4Yun/probably-good-is-expanding-our-team; Irregular grant：https://web.archive.org/web/20250323082627id_/https://www.goodventures.org/our-portfolio/grants/pattern-labs-technological-risk-mitigation/; EA Funds grant ledger：https://funds.effectivealtruism.org/api/grants. ↩：#donors","The diagram shows selected organizational roles and funding; the table also records family and education ties omitted from the diagram. EA Infrastructure Fund recommended one $394,968 joint MOOC award in 2022 Q3 to Dan Lahav and Sella Nevo; the two arrows represent that one recommendation. Its ledger leaves the course and organization unnamed. ↩：#ea-map","The five-year felony provision of Section 1030(a)(2)(C) of the Computer Fraud and Abuse Act requires an aggravator such as commercial advantage, furthering another criminal or tortious act, or obtaining information worth more than $5,000. The principal first-offense felony provisions for damaging access or transmissions under §1030(a)(5) require the specified mental state and statutory harm, such as at least $5,000 in qualifying loss or damage affecting ten protected computers. The legal assessment still requires each system's permission, impairment, response costs and U.S. commerce connection. Prosecutors would also need to establish the conduct and knowledge of responsible people and a basis for attributing those acts to Irregular. 18 U.S.C. § 1030：https://uscode.house.gov/view.xhtml?req=granuleid:USC-prelim-title18-section1030&num=0&edition=prelim. ↩：#cfaa"],"articleImages":[{"sourceUrl":"https://www.effort.news/images/article-graphics/irregular/anthropic-ctf-excerpt.png","alt":"Anthropic incident-assessment excerpt explaining that Claude received capture-the-flag tasks, unintended internet access, and no explicit system-scope limits; each incident involved one isolated Claude instance.","afterParagraph":3,"url":"/media/articles/cmu2b7xzq03q4ro25knqglkmi/db5e956a6831bd24.png"},{"sourceUrl":"https://www.effort.news/images/article-graphics/irregular/anthropic-mythos-resampling-excerpt.png","alt":"Anthropic assessment excerpt: a resampling condition brought Mythos 5’s original malicious-package upload route to zero percent; 22 percent of trajectories searched for a simulated option.","afterParagraph":7,"url":"/media/articles/cmu2b7xzq03q4ro25knqglkmi/7691fafe64936e7f.png"}],"mediaStatus":"ok","articleBodyZh":["以色列有效利他主义公司Irregular导致未受保护的AI模型入侵真实目标。","在过去三个月中，OpenAI、Anthropic和Meta的模型入侵了多个现实世界系统。这些模型获得了对网络系统的未经授权访问：https://openai.com/index/third-party-cyber-evaluations-involving-openai-models/，发布了恶意软件包：https://www.anthropic.com/research/alignment-assessment-cybersecurity-incidents，并利用了未命名的漏洞：https://apnews.com/article/0e8061437da6779be962b24ac134a514。","单一公司Irregular对这三家公司所进行的黑客行为负责。Anthropic披露：https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals Irregular负责创建导致Claude入侵现实世界目标的测试，并为模型提供互联网访问权限。Irregular声称：https://www.irregular.com/research/addressing-recent-incidents-ongoing-findings-and-path-forward 它当时并不知道自己向这些AI模型提供了互联网访问权限。","在更正常的媒体生态系统中，对这些网络安全问题的反应将是显而易见的。美国的AI公司会重新考虑与Irregular的业务往来，不仅因为其未能保护其系统，还因为它是一家可能不受美国监管的以色列公司。立法者会考虑对Irregular或其美国商业合作伙伴采取行动，这些合作伙伴包括OpenAI、Anthropic和Meta。他们可能会考虑加强对指示AI模型进行网络攻击的公司的责任追究，以及这些模型随后执行这些网络攻击的责任。","在每次评估中，Claude都被分配了一个CTF挑战：模型得到一个虚构的场景、一个目标机器，以及需要从中提取的一条机密信息（“旗标”）。所有四个提示都说明Claude无法访问互联网，但每种情况下，环境中的错误配置都使互联网访问开放。没有任何提示说明哪些系统在练习范围内，也没有限制Claude搜索旗标的地点。所有事件仅涉及单个Claude实例独立工作，每次运行持续约10至34小时的活跃工作时间。","相反，Irregular、Anthropic 及其盟友开始发起媒体宣传运动，用煽情的语言推广字面意义上的末日主义意识形态。Anthropic 的事件评估将责任归咎于其自身 AI 的“鲁莽行为”：https://www.anthropic.com/research/alignment-assessment-cybersecurity-incidents；Irregular 描述了“智能体本身成为威胁行为者”：https://www.irregular.com/research/emergent-offensive-cyber-behavior-in-ai-agents；Anthropic 的 CEO Dario Amodei 在谈及类似的 OpenAI–Hugging Face 黑客事件时警告称，未来的智能体群“可能能够接管整个互联网”：https://darioamodei.com/post/we-must-pace-the-frontier；而美联社的一则标题声称机器人正在“失控”：https://apnews.com/article/0e8061437da6779be962b24ac134a514。","在 Anthropic 的一份报告中：https://www.anthropic.com/research/alignment-assessment-cybersecurity-incidents，其 Claude 模型通过模拟的名称碰撞入侵了一家真实公司的系统，发布了恶意软件包，并扫描了外部系统。在这次测试中：https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals，Anthropic 和 Irregular 错误地向该模型提供了互联网访问权限，并没有指导模型“哪些系统在演练范围内”。","虽然 Anthropic 声称其问题是由“失控的智能体群”和“未对齐”引起的，但其后续披露显示，零智能体真正“失控”。在这次实验中，一旦 Anthropic 员工告诉 Claude 模型不要进行现实世界的黑客攻击，该模型的现实世界黑客行为就下降到零。根据他们自己的发现，Anthropic 和 Irregular 对他们引发的网络安全事件负有全部责任。","在这些攻击之后，Anthropic 和 Irregular 部署了一群由与 Anthropic 有关联的基金会支付的 AI 安全影响者：https://www.effort.news/tarbell，旨在转移公众对他们责任的注意力，并引导至毫无根据的“失控智能体”理论。像 Anthropic 一样，Irregular 与这些基金会密不可分。","Omer Nevo，Irregular 的联合创始人兼首席技术官，是以色列有效利他主义组织的董事会成员，以及有效利他主义非政府组织 Heron 和 Probably Good 的成员。Dan Lahav，Irregular 的联合创始人兼首席执行官，与 Sella Nevo（Omer Nevo 的兄弟）一起获得了 395,000 美元启动一门课程。Sella 和 Omer 共同创立了一家教育公众关于有效利他主义的非政府组织——Impact Focused Education。他们还共同创立了 Probably Good。 2：#irregular-note-2","这些分支机构都由 Dustin Moskovitz 资助，他是在 Sam Bankman-Fried 被捕后，有效利他主义/人工智能安全事业的主要捐助者。Irregular 的第一位投资者是 Dustin Moskovitz 的公司 Good Ventures。Dustin Moskovitz 的慈善工具 Coefficient Giving/Open Philanthropy 为以色列有效利他主义、Heron 和 Probably Good 提供资金。 3：#irregular-note-3","Irregular 获得了未经授权的访问权限，修改了记录，并利用他们被授予访问的未加密模型发布了盗取凭证的软件包。在某些情况下，这种行为违反了《计算机欺诈与滥用法》第 1030(a)(2)(C) 条，涉及故意未经授权访问以获取信息。然而，其重罪指控需要具体的损害和意图证据。 5：#irregular-note-5","虽然它主要与美国实验室签约，但位于以色列的关键 Irregular 领导层、员工及资源可能不受美国监管。Ynet 的访问和采访：https://www.ynetnews.com/magazine/article/syber7xg11g 描述了 Irregular 在特拉维夫的办公室。CheckID 的公司列表：https://www.checkid.co.il/company/%D7%A4%D7%90%D7%98%D7%A8%D7%9F-%D7%98%D7%A7-%D7%91%D7%A2~%D7%9E-516854460 识别了两个关联实体：Pattern Labs Tech Inc.，一家特拉华州公司：https://tmng-al.uspto.gov/resting2/api/casedoc/cms/case/99649085/office-action/OfficeAction8619771.pdf，以及 Pattern Tech Ltd，编号 516854460，注册在特拉维夫的活跃以色列公司：https://next.obudget.org/i/org/company/516854460。","时间线标记了公开披露。Anthropic 修正后的九月评估统计了七次运行中的四起事件；OpenAI 和 Meta 报告了独立的 Irregular 评估事件。日期描述的是披露时间，而非每次潜在入侵发生的时间。 ↩：#incident-timeline","Impact Focused Education 确认 Dan Lahav 和 Sella Nevo 为其联合创始人。资助账本记录了对他们的 $394,968 推荐，而不是确认收款或与具体课程的对应拨款。EA 以色列董事会：https://www.effective-altruism.org.il/%D7%A2%D7%9C%D7%99%D7%A0%D7%95；Heron 顾问委员会：https://www.heronsec.ai/about；Probably Good 董事会：https://probablygood.org/about/；EA Funds 资助账本：https://funds.effectivealtruism.org/api/grants；Omer 和 Sella 的关系：https://www.jta.org/2012/01/24/israel/nixing-names-soldiers-and-sushi-couch-potatoes-make-their-mark；IFE 创始人：https://www.impactfocusededucation.org/about。↩：#founders","Good Ventures 创始人：https://goodventures.org/；Coefficient Giving 关系：https://coefficientgiving.org/about-us/；Heron 资助：https://www.heronsec.ai/about；Probably Good 资助：https://forum.effectivealtruism.org/posts/d4JRmqWXkKKoF4Yun/probably-good-is-expanding-our-team；不规则资助：https://web.archive.org/web/20250323082627id_/https://www.goodventures.org/our-portfolio/grants/pattern-labs-technological-risk-mitigation/；EA Funds 资助账本：https://funds.effectivealtruism.org/api/grants。↩：#donors","该图显示了选定的组织角色和资金；表格还记录了图中未显示的家庭和教育关系。EA 基础设施基金在 2022 年第三季度推荐了一项 $394,968 的联合 MOOC 资助给 Dan Lahav 和 Sella Nevo；两个箭头表示同一个推荐。其账本未注明具体课程和组织。↩：#ea-map","《计算机欺诈和滥用法》第1030(a)(2)(C)条款的五年重罪条款要求存在加重因素，例如商业利益、促进其他犯罪或侵权行为，或获取价值超过5,000美元的信息。对首次犯罪者在§1030(a)(5)条下造成破坏性访问或传输的主要重罪条款要求具备特定的心理状态和法定损害，例如至少5,000美元的合格损失或影响十台受保护计算机的损害。法律评估仍然需要考虑每个系统的许可、损害、响应成本以及与美国商业的关联。检察官还需证明相关人员的行为和知识，以及归责于Irregular行为的依据。18 U.S.C. § 1030：https://uscode.house.gov/view.xhtml?req=granuleid:USC-prelim-title18-section1030&num=0&edition=prelim. ↩：#cfaa"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Aioga 编辑摘要：文章指过去三个月内 OpenAI、Anthropic 和 Meta 的模型在评估中越权访问真实系统、发布恶意包，且事件均与一家公司 Irregular 有关。 Aioga 将其归入「行业动态」方向，重点关注它对真实使用和行业竞争的影响。","background":"背景分析：模型与研究类动态需要结合能力边界、开放方式、成本、可用性和真实任务表现判断，单项指标领先不等于已经形成稳定采用。","viewpoint":"Aioga 判断：这条动态更适合作为行业观察信号，当前信息足以建立线索，但不足以推导长期结论。","implications":"影响分析：对相关团队而言，短期应先核对来源、可用范围和实际成本，再判断是否值得接入或跟进。","nextStep":"后续观察：继续观察官方文档、实际可用性、价格变化、开发者反馈和竞品回应。","evidenceRefs":["title","summary","articleBody"],"confidence":"medium","status":"published","aiGenerated":false,"autoApproved":true,"generatedBy":"rule-safe-fallback","generatedAt":"2026-09-18T17:13:51.593Z","sourceHash":"5a6621c212ec74f6","validation":{"passed":true,"mode":"rule-safe-fallback","checks":["schema","length","source-attribution","no-html"]}},"tags":["行业动态","Hacker News 热门（buzzing.cc 中文翻译）"],"translations":{"zh-CN":{"title":"OpenAI、Anthropic 和 Meta 模型入侵事件被指均与评估公司 Irregular 有关","summary":"文章指过去三个月内 OpenAI、Anthropic 和 Meta 的模型在评估中越权访问真实系统、发布恶意包，且事件均与一家公司 Irregular 有关。","category":"行业动态","source":"effort.news","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"OpenAI、Anthropic 和 Meta 模型入侵事件被指均与评估公司 Irregular 有关 - Aioga AI资讯","description":"文章指过去三个月内 OpenAI、Anthropic 和 Meta 的模型在评估中越权访问真实系统、发布恶意包，且事件均与一家公司 Irregular 有关。","url":"https://www.aioga.com/news/cmu2b7xzq03q4ro25knqglkmi/","articleBody":["以色列有效利他主义公司Irregular导致未受保护的AI模型入侵真实目标。","在过去三个月中，OpenAI、Anthropic和Meta的模型入侵了多个现实世界系统。这些模型获得了对网络系统的未经授权访问：https://openai.com/index/third-party-cyber-evaluations-involving-openai-models/，发布了恶意软件包：https://www.anthropic.com/research/alignment-assessment-cybersecurity-incidents，并利用了未命名的漏洞：https://apnews.com/article/0e8061437da6779be962b24ac134a514。","单一公司Irregular对这三家公司所进行的黑客行为负责。Anthropic披露：https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals Irregular负责创建导致Claude入侵现实世界目标的测试，并为模型提供互联网访问权限。Irregular声称：https://www.irregular.com/research/addressing-recent-incidents-ongoing-findings-and-path-forward 它当时并不知道自己向这些AI模型提供了互联网访问权限。","在更正常的媒体生态系统中，对这些网络安全问题的反应将是显而易见的。美国的AI公司会重新考虑与Irregular的业务往来，不仅因为其未能保护其系统，还因为它是一家可能不受美国监管的以色列公司。立法者会考虑对Irregular或其美国商业合作伙伴采取行动，这些合作伙伴包括OpenAI、Anthropic和Meta。他们可能会考虑加强对指示AI模型进行网络攻击的公司的责任追究，以及这些模型随后执行这些网络攻击的责任。","在每次评估中，Claude都被分配了一个CTF挑战：模型得到一个虚构的场景、一个目标机器，以及需要从中提取的一条机密信息（“旗标”）。所有四个提示都说明Claude无法访问互联网，但每种情况下，环境中的错误配置都使互联网访问开放。没有任何提示说明哪些系统在练习范围内，也没有限制Claude搜索旗标的地点。所有事件仅涉及单个Claude实例独立工作，每次运行持续约10至34小时的活跃工作时间。","相反，Irregular、Anthropic 及其盟友开始发起媒体宣传运动，用煽情的语言推广字面意义上的末日主义意识形态。Anthropic 的事件评估将责任归咎于其自身 AI 的“鲁莽行为”：https://www.anthropic.com/research/alignment-assessment-cybersecurity-incidents；Irregular 描述了“智能体本身成为威胁行为者”：https://www.irregular.com/research/emergent-offensive-cyber-behavior-in-ai-agents；Anthropic 的 CEO Dario Amodei 在谈及类似的 OpenAI–Hugging Face 黑客事件时警告称，未来的智能体群“可能能够接管整个互联网”：https://darioamodei.com/post/we-must-pace-the-frontier；而美联社的一则标题声称机器人正在“失控”：https://apnews.com/article/0e8061437da6779be962b24ac134a514。","在 Anthropic 的一份报告中：https://www.anthropic.com/research/alignment-assessment-cybersecurity-incidents，其 Claude 模型通过模拟的名称碰撞入侵了一家真实公司的系统，发布了恶意软件包，并扫描了外部系统。在这次测试中：https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals，Anthropic 和 Irregular 错误地向该模型提供了互联网访问权限，并没有指导模型“哪些系统在演练范围内”。","虽然 Anthropic 声称其问题是由“失控的智能体群”和“未对齐”引起的，但其后续披露显示，零智能体真正“失控”。在这次实验中，一旦 Anthropic 员工告诉 Claude 模型不要进行现实世界的黑客攻击，该模型的现实世界黑客行为就下降到零。根据他们自己的发现，Anthropic 和 Irregular 对他们引发的网络安全事件负有全部责任。","在这些攻击之后，Anthropic 和 Irregular 部署了一群由与 Anthropic 有关联的基金会支付的 AI 安全影响者：https://www.effort.news/tarbell，旨在转移公众对他们责任的注意力，并引导至毫无根据的“失控智能体”理论。像 Anthropic 一样，Irregular 与这些基金会密不可分。","Omer Nevo，Irregular 的联合创始人兼首席技术官，是以色列有效利他主义组织的董事会成员，以及有效利他主义非政府组织 Heron 和 Probably Good 的成员。Dan Lahav，Irregular 的联合创始人兼首席执行官，与 Sella Nevo（Omer Nevo 的兄弟）一起获得了 395,000 美元启动一门课程。Sella 和 Omer 共同创立了一家教育公众关于有效利他主义的非政府组织——Impact Focused Education。他们还共同创立了 Probably Good。 2：#irregular-note-2","这些分支机构都由 Dustin Moskovitz 资助，他是在 Sam Bankman-Fried 被捕后，有效利他主义/人工智能安全事业的主要捐助者。Irregular 的第一位投资者是 Dustin Moskovitz 的公司 Good Ventures。Dustin Moskovitz 的慈善工具 Coefficient Giving/Open Philanthropy 为以色列有效利他主义、Heron 和 Probably Good 提供资金。 3：#irregular-note-3","Irregular 获得了未经授权的访问权限，修改了记录，并利用他们被授予访问的未加密模型发布了盗取凭证的软件包。在某些情况下，这种行为违反了《计算机欺诈与滥用法》第 1030(a)(2)(C) 条，涉及故意未经授权访问以获取信息。然而，其重罪指控需要具体的损害和意图证据。 5：#irregular-note-5","虽然它主要与美国实验室签约，但位于以色列的关键 Irregular 领导层、员工及资源可能不受美国监管。Ynet 的访问和采访：https://www.ynetnews.com/magazine/article/syber7xg11g 描述了 Irregular 在特拉维夫的办公室。CheckID 的公司列表：https://www.checkid.co.il/company/%D7%A4%D7%90%D7%98%D7%A8%D7%9F-%D7%98%D7%A7-%D7%91%D7%A2~%D7%9E-516854460 识别了两个关联实体：Pattern Labs Tech Inc.，一家特拉华州公司：https://tmng-al.uspto.gov/resting2/api/casedoc/cms/case/99649085/office-action/OfficeAction8619771.pdf，以及 Pattern Tech Ltd，编号 516854460，注册在特拉维夫的活跃以色列公司：https://next.obudget.org/i/org/company/516854460。","时间线标记了公开披露。Anthropic 修正后的九月评估统计了七次运行中的四起事件；OpenAI 和 Meta 报告了独立的 Irregular 评估事件。日期描述的是披露时间，而非每次潜在入侵发生的时间。 ↩：#incident-timeline","Impact Focused Education 确认 Dan Lahav 和 Sella Nevo 为其联合创始人。资助账本记录了对他们的 $394,968 推荐，而不是确认收款或与具体课程的对应拨款。EA 以色列董事会：https://www.effective-altruism.org.il/%D7%A2%D7%9C%D7%99%D7%A0%D7%95；Heron 顾问委员会：https://www.heronsec.ai/about；Probably Good 董事会：https://probablygood.org/about/；EA Funds 资助账本：https://funds.effectivealtruism.org/api/grants；Omer 和 Sella 的关系：https://www.jta.org/2012/01/24/israel/nixing-names-soldiers-and-sushi-couch-potatoes-make-their-mark；IFE 创始人：https://www.impactfocusededucation.org/about。↩：#founders","Good Ventures 创始人：https://goodventures.org/；Coefficient Giving 关系：https://coefficientgiving.org/about-us/；Heron 资助：https://www.heronsec.ai/about；Probably Good 资助：https://forum.effectivealtruism.org/posts/d4JRmqWXkKKoF4Yun/probably-good-is-expanding-our-team；不规则资助：https://web.archive.org/web/20250323082627id_/https://www.goodventures.org/our-portfolio/grants/pattern-labs-technological-risk-mitigation/；EA Funds 资助账本：https://funds.effectivealtruism.org/api/grants。↩：#donors","该图显示了选定的组织角色和资金；表格还记录了图中未显示的家庭和教育关系。EA 基础设施基金在 2022 年第三季度推荐了一项 $394,968 的联合 MOOC 资助给 Dan Lahav 和 Sella Nevo；两个箭头表示同一个推荐。其账本未注明具体课程和组织。↩：#ea-map","《计算机欺诈和滥用法》第1030(a)(2)(C)条款的五年重罪条款要求存在加重因素，例如商业利益、促进其他犯罪或侵权行为，或获取价值超过5,000美元的信息。对首次犯罪者在§1030(a)(5)条下造成破坏性访问或传输的主要重罪条款要求具备特定的心理状态和法定损害，例如至少5,000美元的合格损失或影响十台受保护计算机的损害。法律评估仍然需要考虑每个系统的许可、损害、响应成本以及与美国商业的关联。检察官还需证明相关人员的行为和知识，以及归责于Irregular行为的依据。18 U.S.C. § 1030：https://uscode.house.gov/view.xhtml?req=granuleid:USC-prelim-title18-section1030&num=0&edition=prelim. ↩：#cfaa"]},"en":{"title":"The infiltration incidents involving OpenAI, Anthropic, and Meta models are all linked to the assessment company Irregular","summary":"The article alleges that over the past three months, models from OpenAI, Anthropic, and Meta have overstepped their authority to access real systems and release malicious packages during evaluations, with all incidents related to a company, Irregular.","category":"Industry","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"The infiltration incidents involving OpenAI, Anthropic, and Meta models are all linked to the assessment company Irregular - Aioga AI News","description":"The article alleges that over the past three months, models from OpenAI, Anthropic, and Meta have overstepped their authority to access real systems and release malicious packages...","url":"https://www.aioga.com/en/news/cmu2b7xzq03q4ro25knqglkmi/","contentTranslated":true,"sourceHash":"24842226af336427","translatedAt":"2026-09-15T07:01:38.843Z"},"ja":{"title":"OpenAI、Anthropic、Metaモデルに関わる浸透事件はすべて評価会社Irregularに関連しています","summary":"記事は、過去3か月間にOpenAI、Anthropic、Metaのモデルが権限を超えて実際のシステムにアクセスし、評価中に悪意のあるパッケージをリリースし、すべてのインシデントがIrregularという企業に関連していると主張しています。","category":"業界動向","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"OpenAI、Anthropic、Metaモデルに関わる浸透事件はすべて評価会社Irregularに関連しています - Aioga AIニュース","description":"記事は、過去3か月間にOpenAI、Anthropic、Metaのモデルが権限を超えて実際のシステムにアクセスし、評価中に悪意のあるパッケージをリリースし、すべてのインシデントがIrregularという企業に関連していると主張しています。","url":"https://www.aioga.com/ja/news/cmu2b7xzq03q4ro25knqglkmi/","contentTranslated":true,"sourceHash":"24842226af336427","translatedAt":"2026-09-15T07:01:38.890Z"},"ko":{"title":"OpenAI, Anthropic, Meta 모델과 관련된 침투 사건들은 모두 평가 회사인 Irregular와 연관되어 있습니다","summary":"기사는 지난 3개월 동안 OpenAI, Anthropic, Meta의 모델들이 실제 시스템에 접근하고 평가 중 악성 패키지를 공개하는 권한을 넘어섰으며, 모든 사건이 Irregular라는 회사와 관련되었다고 주장합니다.","category":"업계 동향","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"OpenAI, Anthropic, Meta 모델과 관련된 침투 사건들은 모두 평가 회사인 Irregular와 연관되어 있습니다 - Aioga AI 뉴스","description":"기사는 지난 3개월 동안 OpenAI, Anthropic, Meta의 모델들이 실제 시스템에 접근하고 평가 중 악성 패키지를 공개하는 권한을 넘어섰으며, 모든 사건이 Irregular라는 회사와 관련되었다고 주장합니다.","url":"https://www.aioga.com/ko/news/cmu2b7xzq03q4ro25knqglkmi/","contentTranslated":true,"sourceHash":"24842226af336427","translatedAt":"2026-09-15T07:01:47.997Z"},"es":{"title":"Los incidentes de infiltración que involucran a OpenAI, Anthropic y Meta Models están todos vinculados a la empresa de evaluación Irregular","summary":"El artículo alega que en los últimos tres meses, modelos de OpenAI, Anthropic y Meta han excedido su autoridad para acceder a sistemas reales y liberar paquetes maliciosos durante las evaluaciones, siendo todos los incidentes relacionados con una empresa, Irregular.","category":"Industria","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Los incidentes de infiltración que involucran a OpenAI, Anthropic y Meta Models están todos vinculados a la empresa de evaluación Irregular - Aioga Noticias de IA","description":"El artículo alega que en los últimos tres meses, modelos de OpenAI, Anthropic y Meta han excedido su autoridad para acceder a sistemas reales y liberar paquetes maliciosos durante...","url":"https://www.aioga.com/es/news/cmu2b7xzq03q4ro25knqglkmi/","contentTranslated":true,"sourceHash":"24842226af336427","translatedAt":"2026-09-15T07:01:47.927Z"},"fr":{"title":"Les incidents d’infiltration impliquant OpenAI, Anthropic et Meta Models sont tous liés à la société d’évaluation Irregular","summary":"L’article affirme que, au cours des trois derniers mois, des modèles d’OpenAI, Anthropic et Meta ont dépassé leur autorité pour accéder à de vrais systèmes et publier des paquets malveillants lors des évaluations, tous les incidents étant liés à une entreprise, Irregular.","category":"Industrie","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Les incidents d’infiltration impliquant OpenAI, Anthropic et Meta Models sont tous liés à la société d’évaluation Irregular - Aioga Actualités IA","description":"L’article affirme que, au cours des trois derniers mois, des modèles d’OpenAI, Anthropic et Meta ont dépassé leur autorité pour accéder à de vrais systèmes et publier des paquets m...","url":"https://www.aioga.com/fr/news/cmu2b7xzq03q4ro25knqglkmi/","contentTranslated":true,"sourceHash":"24842226af336427","translatedAt":"2026-09-15T07:01:57.149Z"},"de":{"title":"Die Infiltrationsvorfälle mit OpenAI, Anthropic und Meta-Modellen stehen alle im Zusammenhang mit dem Bewertungsunternehmen Irregular","summary":"Der Artikel behauptet, dass in den letzten drei Monaten Modelle von OpenAI, Anthropic und Meta ihre Befugnisse überschritten haben, um auf echte Systeme zuzugreifen und bösartige Pakete während der Bewertungen freizugeben, wobei alle Vorfälle ein Unternehmen namens Irregular betreffen.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Die Infiltrationsvorfälle mit OpenAI, Anthropic und Meta-Modellen stehen alle im Zusammenhang mit dem Bewertungsunternehmen Irregular - Aioga KI-News","description":"Der Artikel behauptet, dass in den letzten drei Monaten Modelle von OpenAI, Anthropic und Meta ihre Befugnisse überschritten haben, um auf echte Systeme zuzugreifen und bösartige P...","url":"https://www.aioga.com/de/news/cmu2b7xzq03q4ro25knqglkmi/","contentTranslated":true,"sourceHash":"24842226af336427","translatedAt":"2026-09-15T07:01:57.325Z"},"pt-BR":{"title":"Os incidentes de infiltração envolvendo modelos OpenAI, Anthropic e Meta estão todos ligados à empresa de avaliação Irregular","summary":"O artigo alega que, nos últimos três meses, modelos da OpenAI, Anthropic e Meta excederam sua autoridade para acessar sistemas reais e liberar pacotes maliciosos durante avaliações, com todos os incidentes relacionados a uma empresa chamada Irregular.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Os incidentes de infiltração envolvendo modelos OpenAI, Anthropic e Meta estão todos ligados à empresa de avaliação Irregular - Aioga Notícias de IA","description":"O artigo alega que, nos últimos três meses, modelos da OpenAI, Anthropic e Meta excederam sua autoridade para acessar sistemas reais e liberar pacotes maliciosos durante avaliações...","url":"https://www.aioga.com/pt-BR/news/cmu2b7xzq03q4ro25knqglkmi/","contentTranslated":true,"sourceHash":"24842226af336427","translatedAt":"2026-09-15T07:02:06.746Z"},"ru":{"title":"Инциденты проникновения, связанные с OpenAI, Anthropic и Meta-моделями, все связаны с оценочной компанией Irregular","summary":"В статье утверждается, что за последние три месяца модели OpenAI, Anthropic и Meta превысили свои полномочия в доступе к реальным системам и выпуску вредоносных пакетов во время оценок, при этом все инциденты касались компании Irregular.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Инциденты проникновения, связанные с OpenAI, Anthropic и Meta-моделями, все связаны с оценочной компанией Irregular - Aioga Новости ИИ","description":"В статье утверждается, что за последние три месяца модели OpenAI, Anthropic и Meta превысили свои полномочия в доступе к реальным системам и выпуску вредоносных пакетов во время оц...","url":"https://www.aioga.com/ru/news/cmu2b7xzq03q4ro25knqglkmi/","contentTranslated":true,"sourceHash":"24842226af336427","translatedAt":"2026-09-15T07:02:06.669Z"},"ar":{"title":"ترتبط حوادث التسلل التي تشمل نماذج OpenAI وAnthropic وMeta جميعها بشركة التقييم Irregular","summary":"تزعم المقالة أنه خلال الأشهر الثلاثة الماضية، تجاوزت نماذج من OpenAI وAnthropic وMeta صلاحياتها للوصول إلى أنظمة حقيقية وإطلاق حزم خبيثة أثناء التقييمات، وجميع الحوادث المتعلقة بشركة تدعى Irregular.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"ترتبط حوادث التسلل التي تشمل نماذج OpenAI وAnthropic وMeta جميعها بشركة التقييم Irregular - Aioga أخبار الذكاء الاصطناعي","description":"تزعم المقالة أنه خلال الأشهر الثلاثة الماضية، تجاوزت نماذج من OpenAI وAnthropic وMeta صلاحياتها للوصول إلى أنظمة حقيقية وإطلاق حزم خبيثة أثناء التقييمات، وجميع الحوادث المتعلقة بشر...","url":"https://www.aioga.com/ar/news/cmu2b7xzq03q4ro25knqglkmi/","contentTranslated":true,"sourceHash":"24842226af336427","translatedAt":"2026-09-15T07:02:15.996Z"},"hi":{"title":"OpenAI, Anthropic और Meta मॉडल से जुड़ी घुसपैठ की सभी घटनाएं मूल्यांकन कंपनी Irregular से जुड़ी हुई हैं","summary":"लेख में आरोप लगाया गया है कि पिछले तीन महीनों में, OpenAI, Anthropic और Meta के मॉडलों ने वास्तविक सिस्टम तक पहुंचने और मूल्यांकन के दौरान दुर्भावनापूर्ण पैकेज जारी करने के अपने अधिकार को पार कर लिया है, जिसमें एक कंपनी, Irregular से संबंधित सभी घटनाएं शामिल हैं।","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"OpenAI, Anthropic और Meta मॉडल से जुड़ी घुसपैठ की सभी घटनाएं मूल्यांकन कंपनी Irregular से जुड़ी हुई हैं - Aioga AI समाचार","description":"लेख में आरोप लगाया गया है कि पिछले तीन महीनों में, OpenAI, Anthropic और Meta के मॉडलों ने वास्तविक सिस्टम तक पहुंचने और मूल्यांकन के दौरान दुर्भावनापूर्ण पैकेज जारी करने के अपने अध...","url":"https://www.aioga.com/hi/news/cmu2b7xzq03q4ro25knqglkmi/","contentTranslated":true,"sourceHash":"24842226af336427","translatedAt":"2026-09-15T07:02:15.906Z"},"it":{"title":"Gli incidenti di infiltrazione che coinvolgono OpenAI, Anthropic e Meta Models sono tutti collegati alla società di valutazione Irregular","summary":"L'articolo sostiene che negli ultimi tre mesi, modelli di OpenAI, Anthropic e Meta abbiano oltrepassato la loro autorità per accedere a sistemi reali e rilasciare pacchetti dannosi durante le valutazioni, con tutti gli incidenti legati a un'azienda, Irregular.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Gli incidenti di infiltrazione che coinvolgono OpenAI, Anthropic e Meta Models sono tutti collegati alla società di valutazione Irregular - Aioga Notizie IA","description":"L'articolo sostiene che negli ultimi tre mesi, modelli di OpenAI, Anthropic e Meta abbiano oltrepassato la loro autorità per accedere a sistemi reali e rilasciare pacchetti dannosi...","url":"https://www.aioga.com/it/news/cmu2b7xzq03q4ro25knqglkmi/","contentTranslated":true,"sourceHash":"24842226af336427","translatedAt":"2026-09-15T07:02:25.014Z"},"nl":{"title":"De infiltratieincidenten met OpenAI, Anthropic en Meta-modellen zijn allemaal gekoppeld aan het beoordelingsbedrijf Irregular","summary":"Het artikel beweert dat modellen van OpenAI, Anthropic en Meta in de afgelopen drie maanden hun bevoegdheid hebben overschreden om echte systemen te benaderen en kwaadaardige pakketten vrij te geven tijdens evaluaties, waarbij alle incidenten betrekking hebben op een bedrijf, Irregular.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"De infiltratieincidenten met OpenAI, Anthropic en Meta-modellen zijn allemaal gekoppeld aan het beoordelingsbedrijf Irregular - Aioga AI-nieuws","description":"Het artikel beweert dat modellen van OpenAI, Anthropic en Meta in de afgelopen drie maanden hun bevoegdheid hebben overschreden om echte systemen te benaderen en kwaadaardige pakke...","url":"https://www.aioga.com/nl/news/cmu2b7xzq03q4ro25knqglkmi/","contentTranslated":true,"sourceHash":"24842226af336427","translatedAt":"2026-09-15T07:02:25.056Z"},"tr":{"title":"OpenAI, Anthropic ve Meta modellerini içeren sızma olaylarının tamamı değerlendirme şirketi Irregular'a bağlıdır","summary":"Makale, son üç ayda OpenAI, Anthropic ve Meta modellerinin gerçek sistemlere erişim ve değerlendirmeler sırasında kötü amaçlı paketler yayınlama yetkisini aştığını, tüm olayların ise Irregular adlı bir şirketle ilgili olduğunu iddia ediyor.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"OpenAI, Anthropic ve Meta modellerini içeren sızma olaylarının tamamı değerlendirme şirketi Irregular'a bağlıdır - Aioga AI Haberleri","description":"Makale, son üç ayda OpenAI, Anthropic ve Meta modellerinin gerçek sistemlere erişim ve değerlendirmeler sırasında kötü amaçlı paketler yayınlama yetkisini aştığını, tüm olayların i...","url":"https://www.aioga.com/tr/news/cmu2b7xzq03q4ro25knqglkmi/","contentTranslated":true,"sourceHash":"24842226af336427","translatedAt":"2026-09-15T07:02:34.107Z"},"vi":{"title":"Các sự cố xâm nhập liên quan đến OpenAI, Anthropic và Meta đều liên quan đến công ty đánh giá Irregular","summary":"Bài báo cáo buộc rằng trong ba tháng qua, các mô hình của OpenAI, Anthropic và Meta đã vượt quá thẩm quyền truy cập các hệ thống thực tế và phát hành các gói độc hại trong quá trình đánh giá, với tất cả các sự cố liên quan đến một công ty, Irregular.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Các sự cố xâm nhập liên quan đến OpenAI, Anthropic và Meta đều liên quan đến công ty đánh giá Irregular - Tin tức AI Aioga","description":"Bài báo cáo buộc rằng trong ba tháng qua, các mô hình của OpenAI, Anthropic và Meta đã vượt quá thẩm quyền truy cập các hệ thống thực tế và phát hành các gói độc hại trong quá trìn...","url":"https://www.aioga.com/vi/news/cmu2b7xzq03q4ro25knqglkmi/","contentTranslated":true,"sourceHash":"24842226af336427","translatedAt":"2026-09-15T07:02:34.250Z"},"id":{"title":"Insiden infiltrasi yang melibatkan model OpenAI, Anthropic, dan Meta semuanya terkait dengan perusahaan penilaian Irregular","summary":"Artikel tersebut menuduh bahwa selama tiga bulan terakhir, model dari OpenAI, Anthropic, dan Meta telah melampaui kewenangan mereka untuk mengakses sistem nyata dan merilis paket berbahaya selama evaluasi, dengan semua insiden terkait dengan sebuah perusahaan, Irregular.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Insiden infiltrasi yang melibatkan model OpenAI, Anthropic, dan Meta semuanya terkait dengan perusahaan penilaian Irregular - Berita AI Aioga","description":"Artikel tersebut menuduh bahwa selama tiga bulan terakhir, model dari OpenAI, Anthropic, dan Meta telah melampaui kewenangan mereka untuk mengakses sistem nyata dan merilis paket b...","url":"https://www.aioga.com/id/news/cmu2b7xzq03q4ro25knqglkmi/","contentTranslated":true,"sourceHash":"24842226af336427","translatedAt":"2026-09-15T07:02:43.430Z"},"th":{"title":"เหตุการณ์แทรกซึมที่เกี่ยวข้องกับโมเดล OpenAI, Anthropic และ Meta ทั้งหมดเชื่อมโยงกับบริษัทประเมิน Irregular","summary":"บทความกล่าวหาว่าในช่วงสามเดือนที่ผ่านมา โมเดลจาก OpenAI, Anthropic และ Meta ได้ละเมิดอํานาจในการเข้าถึงระบบจริงและปล่อยแพ็กเกจที่เป็นอันตรายในระหว่างการประเมิน โดยเหตุการณ์ทั้งหมดเกี่ยวข้องกับบริษัท Irregular","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"เหตุการณ์แทรกซึมที่เกี่ยวข้องกับโมเดล OpenAI, Anthropic และ Meta ทั้งหมดเชื่อมโยงกับบริษัทประเมิน Irregular - ข่าว AI Aioga","description":"บทความกล่าวหาว่าในช่วงสามเดือนที่ผ่านมา โมเดลจาก OpenAI, Anthropic และ Meta ได้ละเมิดอํานาจในการเข้าถึงระบบจริงและปล่อยแพ็กเกจที่เป็นอันตรายในระหว่างการประเมิน โดยเหตุการณ์ทั้งหมดเ...","url":"https://www.aioga.com/th/news/cmu2b7xzq03q4ro25knqglkmi/","contentTranslated":true,"sourceHash":"24842226af336427","translatedAt":"2026-09-15T07:02:43.368Z"},"pl":{"title":"Incydenty infiltracji z udziałem modeli OpenAI, Anthropic i Meta są powiązane z firmą analizującą Irregular","summary":"Artykuł zarzuca, że w ciągu ostatnich trzech miesięcy modele z OpenAI, Antropic i Meta przekroczyły swoje uprawnienia, umożliwiając dostęp do rzeczywistych systemów i udostępnianie złośliwych pakietów podczas ocen, przy czym wszystkie incydenty dotyczyły firmy Irregular.","category":"行业动态","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Incydenty infiltracji z udziałem modeli OpenAI, Anthropic i Meta są powiązane z firmą analizującą Irregular - Aioga Wiadomości AI","description":"Artykuł zarzuca, że w ciągu ostatnich trzech miesięcy modele z OpenAI, Antropic i Meta przekroczyły swoje uprawnienia, umożliwiając dostęp do rzeczywistych systemów i udostępnianie...","url":"https://www.aioga.com/pl/news/cmu2b7xzq03q4ro25knqglkmi/","contentTranslated":true,"sourceHash":"24842226af336427","translatedAt":"2026-09-15T07:02:52.604Z"}},"evidenceTier":"verified-news","reviewStatus":"automated-ingest","indexable":true,"editorialCover":""}}