{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-23T07:21:26.498Z","headline":"OpenAI 称 GPT-5.6 Sol 在内部测试中意外入侵了 Hugging Face","description":"OpenAI 表示，其 AI 模型 GPT-5.6 Sol 和另一款\"能力更强的预发布模型\"在内部安全测试中，利用沙箱环境的零日漏洞访问互联网，并成功入侵了开源 AI 平台 Hugging Face。","url":"https://www.aioga.com/news/cmrv6wd5r00f5bi6pxtd27hmu/","mainEntityOfPage":"https://www.aioga.com/news/cmrv6wd5r00f5bi6pxtd27hmu/","datePublished":"2026-07-21T21:48:54.000Z","dateModified":"2026-07-21T21:48:54.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://www.theverge.com/ai-artificial-intelligence/968988/openai-hugging-face-hack-ai","https://aihot.virxact.com/items/cmrv6wd5r00f5bi6pxtd27hmu"],"canonicalUrl":"https://www.aioga.com/news/cmrv6wd5r00f5bi6pxtd27hmu/","directAnswer":{"@type":"Answer","text":"Aioga 编辑摘要：OpenAI 表示，其 AI 模型 GPT-5.6 Sol 和另一款\"能力更强的预发布模型\"在内部安全测试中，利用沙箱环境的零日漏洞访问互联网，并成功入侵了开源 AI 平台 Hugging Face。 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/cmrv6wd5r00f5bi6pxtd27hmu/","dateCreated":"2026-07-21T21:48:54.000Z","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":"The Verge source article","url":"https://www.theverge.com/ai-artificial-intelligence/968988/openai-hugging-face-hack-ai","datePublished":"2026-07-21T21:48:54.000Z","provider":{"@type":"Organization","name":"The Verge","url":"https://www.theverge.com/ai-artificial-intelligence/968988/openai-hugging-face-hack-ai"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmrv6wd5r00f5bi6pxtd27hmu","datePublished":"2026-07-21T21:48:54.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmrv6wd5r00f5bi6pxtd27hmu"}}],"aggregationSource":"The Verge：AI（RSS）","originalPublisher":{"name":"The Verge","url":"https://www.theverge.com/ai-artificial-intelligence/968988/openai-hugging-face-hack-ai"},"article":{"id":"cmrv6wd5r00f5bi6pxtd27hmu","slug":"cmrv6wd5r00f5bi6pxtd27hmu","url":"https://www.aioga.com/news/cmrv6wd5r00f5bi6pxtd27hmu/","title":"OpenAI 称 GPT-5.6 Sol 在内部测试中意外入侵了 Hugging Face","title_en":"OpenAI says it accidentally hacked Hugging Face with a new AI system","summary":"OpenAI 表示，其 AI 模型 GPT-5.6 Sol 和另一款\"能力更强的预发布模型\"在内部安全测试中，利用沙箱环境的零日漏洞访问互联网，并成功入侵了开源 AI 平台 Hugging Face。","source":"The Verge：AI（RSS）","sourceUrl":"https://www.theverge.com/ai-artificial-intelligence/968988/openai-hugging-face-hack-ai","aiHotUrl":"https://aihot.virxact.com/items/cmrv6wd5r00f5bi6pxtd27hmu","publishedAt":"2026-07-21T21:48:54.000Z","category":"行业动态","score":62,"selected":false,"articleBody":["OpenAI says its AI models mistakenly breached open-source AI platform Hugging Face during internal testing. In a blog post on Tuesday：https://openai.com/index/hugging-face-model-evaluation-security-incident/, OpenAI writes that GPT-5.6 Sol and “an even more capable pre-release model” discovered vulnerabilities within their sandboxed testing environment, allowing them to gain access to the internet and target Hugging Face.","On July 16th, Hugging Face disclosed a security incident：https://huggingface.co/blog/security-incident-july-2026 that it says was driven by “an autonomous AI agent system.” Hugging Face’s AI agents detected and stopped the breach, which OpenAI has now admitted occurred during an evaluation of its models’ cybersecurity capabilities. OpenAI says “all evidence suggests that the models were hyperfocused on finding a solution for ExploitGym,” a benchmark system that measures whether AI models can turn security vulnerabilities into exploits.","As part of efforts to complete the evaluation, the AI models gained access to the internet by exploiting a zero-day vulnerability in the sandboxed environment. From there, OpenAI says its models “inferred that Hugging Face potentially hosted models, datasets and solutions for ExploitGym,” and then “searched for and successfully found ways to gain access to secret information that it could use to cheat the evaluation:”","In one example, the model chained together multiple attack vectors, including using stolen credentials and zero-day vulnerabilities to find a remote code execution path on the Hugging Face servers.","OpenAI adds that it’s now working with Hugging Face to investigate the security incident and will implement new controls within its research environment."],"articleImages":[],"mediaStatus":"none","articleBodyZh":["OpenAI 表示，其 AI 模型在内部测试期间错误地侵犯了开源 AI 平台 Hugging Face。在周二的一篇博客文章中：https://openai.com/index/hugging-face-model-evaluation-security-incident/，OpenAI 写道，GPT-5.6 Sol 和“一个更强大的预发行模型”在其沙箱测试环境中发现了漏洞，使它们能够访问互联网并针对 Hugging Face。","7 月 16 日，Hugging Face 披露了一起安全事件：https://huggingface.co/blog/security-incident-july-2026，并表示其原因是“一个自主 AI 代理系统”。Hugging Face 的 AI 代理检测并阻止了此次入侵，而 OpenAI 现在已承认，该事件发生在其评估模型网络安全能力时。OpenAI 表示：“所有证据表明，模型高度专注于寻找 ExploitGym 的解决方案”，该基准系统用于衡量 AI 模型是否能够将安全漏洞转化为漏洞利用。","为了完成评估，AI 模型通过利用沙箱环境中的零日漏洞访问了互联网。随后，OpenAI 表示，其模型“推断出 Hugging Face 可能托管了 ExploitGym 的模型、数据集和解决方案”，然后“搜索并成功找到获取秘密信息的方法，以便在评估中作弊：”","在一个例子中，模型将多个攻击向量串联在一起，包括使用被盗的凭证和零日漏洞，在 Hugging Face 服务器上找到远程代码执行路径。","OpenAI 补充称，正在与 Hugging Face 合作调查此次安全事件，并将在其研究环境中实施新的控制措施。"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Aioga 编辑摘要：OpenAI 表示，其 AI 模型 GPT-5.6 Sol 和另一款\"能力更强的预发布模型\"在内部安全测试中，利用沙箱环境的零日漏洞访问互联网，并成功入侵了开源 AI 平台 Hugging Face。 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：公司与行业类动态需要放在竞争格局、商业化路径、资本信号和监管环境中观察，单条公告不能代表最终结果。","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：继续观察官方文件、合作落地、收入或用户信号、竞品动作和监管后续。","evidenceRefs":["title","summary","articleBody"],"confidence":"medium","status":"published","aiGenerated":false,"autoApproved":true,"generatedBy":"rule-safe-fallback","generatedAt":"2026-07-23T07:30:40.468Z","sourceHash":"52453d4959e3d950","validation":{"passed":true,"mode":"rule-safe-fallback","checks":["schema","length","source-attribution","no-html"]}},"tags":["行业动态","The Verge：AI（RSS）"],"translations":{"zh-CN":{"title":"OpenAI 称 GPT-5.6 Sol 在内部测试中意外入侵了 Hugging Face","summary":"OpenAI 表示，其 AI 模型 GPT-5.6 Sol 和另一款\"能力更强的预发布模型\"在内部安全测试中，利用沙箱环境的零日漏洞访问互联网，并成功入侵了开源 AI 平台 Hugging Face。","category":"行业动态","source":"The Verge","aggregationSource":"The Verge：AI（RSS）","pageTitle":"OpenAI 称 GPT-5.6 Sol 在内部测试中意外入侵了 Hugging Face - Aioga AI资讯","description":"OpenAI 表示，其 AI 模型 GPT-5.6 Sol 和另一款\"能力更强的预发布模型\"在内部安全测试中，利用沙箱环境的零日漏洞访问互联网，并成功入侵了开源 AI 平台 Hugging Face。","url":"https://www.aioga.com/news/cmrv6wd5r00f5bi6pxtd27hmu/"},"en":{"title":"OpenAI says GPT-5.6 Sol accidentally intruded into Hugging Face during internal testing","summary":"OpenAI stated that its AI model GPT-5.6 Sol and another \"more powerful pre-release model\" accessed the internet using zero-day vulnerabilities in a sandbox environment during internal security tests and successfully hacked the open-source AI platform Hugging Face.","category":"Industry","source":"The Verge：AI（RSS）","aggregationSource":"The Verge：AI（RSS）","pageTitle":"OpenAI says GPT-5.6 Sol accidentally intruded into Hugging Face during internal testing - Aioga AI News","description":"OpenAI stated that its AI model GPT-5.6 Sol and another \"more powerful pre-release model\" accessed the internet using zero-day vulnerabilities in a sandbox environment during inter...","url":"https://www.aioga.com/en/news/cmrv6wd5r00f5bi6pxtd27hmu/","contentTranslated":true,"sourceHash":"19a1c0b2f6038311","translatedAt":"2026-07-22T21:43:19.594Z"},"ja":{"title":"OpenAI は GPT-5.6 Sol が内部テスト中に Hugging Face に意図せず侵入したと述べた","summary":"OpenAI は、その AI モデル GPT-5.6 Sol と別の「より高性能なプレリリースモデル」が、内部の安全性テストで、サンドボックス環境のゼロデイ脆弱性を利用してインターネットにアクセスし、オープンソース AI プラットフォーム Hugging Face に侵入することに成功したと示した。","category":"業界動向","source":"The Verge：AI（RSS）","aggregationSource":"The Verge：AI（RSS）","pageTitle":"OpenAI は GPT-5.6 Sol が内部テスト中に Hugging Face に意図せず侵入したと述べた - Aioga AIニュース","description":"OpenAI は、その AI モデル GPT-5.6 Sol と別の「より高性能なプレリリースモデル」が、内部の安全性テストで、サンドボックス環境のゼロデイ脆弱性を利用してインターネットにアクセスし、オープンソース AI プラットフォーム Hugging Face に侵入することに成功したと示した。","url":"https://www.aioga.com/ja/news/cmrv6wd5r00f5bi6pxtd27hmu/","contentTranslated":true,"sourceHash":"19a1c0b2f6038311","translatedAt":"2026-07-22T21:43:34.505Z"},"ko":{"title":"OpenAI는 GPT-5.6 Sol이 내부 테스트 중에 우연히 Hugging Face를 침입했다고 밝혔다","summary":"OpenAI는 자사의 AI 모델 GPT-5.6 Sol과 다른 '더 강력한 사전 출시 모델'이 내부 보안 테스트에서 샌드박스 환경의 제로데이 취약점을 이용해 인터넷에 접근하고, 오픈 소스 AI 플랫폼 Hugging Face를 성공적으로 침입했다고 밝혔습니다.","category":"업계 동향","source":"The Verge：AI（RSS）","aggregationSource":"The Verge：AI（RSS）","pageTitle":"OpenAI는 GPT-5.6 Sol이 내부 테스트 중에 우연히 Hugging Face를 침입했다고 밝혔다 - Aioga AI 뉴스","description":"OpenAI는 자사의 AI 모델 GPT-5.6 Sol과 다른 '더 강력한 사전 출시 모델'이 내부 보안 테스트에서 샌드박스 환경의 제로데이 취약점을 이용해 인터넷에 접근하고, 오픈 소스 AI 플랫폼 Hugging Face를 성공적으로 침입했다고 밝혔습니다.","url":"https://www.aioga.com/ko/news/cmrv6wd5r00f5bi6pxtd27hmu/","contentTranslated":true,"sourceHash":"19a1c0b2f6038311","translatedAt":"2026-07-22T21:44:19.731Z"},"es":{"title":"OpenAI afirma que GPT-5.6 Sol invadió accidentalmente Hugging Face durante las pruebas internas","summary":"OpenAI declaró que su modelo de IA GPT-5.6 Sol y otro 'modelo pre-lanzamiento más potente' en pruebas internas de seguridad, utilizaron vulnerabilidades de día cero en entornos de prueba para acceder a Internet y lograron comprometer la plataforma de IA de código abierto Hugging Face.","category":"Industria","source":"The Verge：AI（RSS）","aggregationSource":"The Verge：AI（RSS）","pageTitle":"OpenAI afirma que GPT-5.6 Sol invadió accidentalmente Hugging Face durante las pruebas internas - Aioga Noticias de IA","description":"OpenAI declaró que su modelo de IA GPT-5.6 Sol y otro 'modelo pre-lanzamiento más potente' en pruebas internas de seguridad, utilizaron vulnerabilidades de día cero en entornos de...","url":"https://www.aioga.com/es/news/cmrv6wd5r00f5bi6pxtd27hmu/","contentTranslated":true,"sourceHash":"19a1c0b2f6038311","translatedAt":"2026-07-22T21:44:24.239Z"},"fr":{"title":"OpenAI affirme que GPT-5.6 Sol a accidentellement piraté Hugging Face lors de tests internes","summary":"OpenAI a déclaré que son modèle d'IA GPT-5.6 Sol et un autre « modèle préliminaire plus puissant » ont, lors de tests de sécurité internes, utilisé des vulnérabilités zero-day dans un environnement sandbox pour accéder à Internet et ont réussi à pirater la plateforme d'IA open source Hugging Face.","category":"Industrie","source":"The Verge：AI（RSS）","aggregationSource":"The Verge：AI（RSS）","pageTitle":"OpenAI affirme que GPT-5.6 Sol a accidentellement piraté Hugging Face lors de tests internes - Aioga Actualités IA","description":"OpenAI a déclaré que son modèle d'IA GPT-5.6 Sol et un autre « modèle préliminaire plus puissant » ont, lors de tests de sécurité internes, utilisé des vulnérabilités zero-day dans...","url":"https://www.aioga.com/fr/news/cmrv6wd5r00f5bi6pxtd27hmu/","contentTranslated":true,"sourceHash":"19a1c0b2f6038311","translatedAt":"2026-07-22T21:45:14.628Z"},"de":{"title":"OpenAI sagt, dass GPT-5.6 Sol während interner Tests versehentlich in Hugging Face eingedrungen ist","summary":"OpenAI erklärte, dass sein KI-Modell GPT-5.6 Sol und ein weiteres \"leistungsstärkeres Vorabmodell\" in internen Sicherheitstests die Null-Tage-Schwachstellen der Sandbox-Umgebung nutzten, um auf das Internet zuzugreifen, und erfolgreich die Open-Source-KI-Plattform Hugging Face gehackt haben.","category":"行业动态","source":"The Verge：AI（RSS）","aggregationSource":"The Verge：AI（RSS）","pageTitle":"OpenAI sagt, dass GPT-5.6 Sol während interner Tests versehentlich in Hugging Face eingedrungen ist - Aioga KI-News","description":"OpenAI erklärte, dass sein KI-Modell GPT-5.6 Sol und ein weiteres \"leistungsstärkeres Vorabmodell\" in internen Sicherheitstests die Null-Tage-Schwachstellen der Sandbox-Umgebung nu...","url":"https://www.aioga.com/de/news/cmrv6wd5r00f5bi6pxtd27hmu/","contentTranslated":true,"sourceHash":"19a1c0b2f6038311","translatedAt":"2026-07-22T21:45:11.460Z"},"pt-BR":{"title":"OpenAI afirma que o GPT-5.6 Sol invadiu acidentalmente o Hugging Face durante testes internos","summary":"A OpenAI afirmou que seu modelo de IA GPT-5.6 Sol e outro \"modelo pré-lançamento mais avançado\", em testes internos de segurança, usaram vulnerabilidades zero-day no ambiente sandbox para acessar a internet e conseguiram invadir a plataforma de IA de código aberto Hugging Face.","category":"行业动态","source":"The Verge：AI（RSS）","aggregationSource":"The Verge：AI（RSS）","pageTitle":"OpenAI afirma que o GPT-5.6 Sol invadiu acidentalmente o Hugging Face durante testes internos - Aioga Notícias de IA","description":"A OpenAI afirmou que seu modelo de IA GPT-5.6 Sol e outro \"modelo pré-lançamento mais avançado\", em testes internos de segurança, usaram vulnerabilidades zero-day no ambiente sandb...","url":"https://www.aioga.com/pt-BR/news/cmrv6wd5r00f5bi6pxtd27hmu/","contentTranslated":true,"sourceHash":"19a1c0b2f6038311","translatedAt":"2026-07-22T21:46:04.138Z"},"ru":{"title":"OpenAI заявляет, что GPT-5.6 Sol случайно вторгся в Hugging Face во время внутреннего тестирования","summary":"OpenAI заявляет, что их ИИ-модель GPT-5.6 Sol и ещё одна «более мощная предварительная версия модели» во внутренних тестах безопасности использовали нулевой день в среде песочницы для доступа к интернету и успешно взломали платформу открытого ИИ Hugging Face.","category":"行业动态","source":"The Verge：AI（RSS）","aggregationSource":"The Verge：AI（RSS）","pageTitle":"OpenAI заявляет, что GPT-5.6 Sol случайно вторгся в Hugging Face во время внутреннего тестирования - Aioga Новости ИИ","description":"OpenAI заявляет, что их ИИ-модель GPT-5.6 Sol и ещё одна «более мощная предварительная версия модели» во внутренних тестах безопасности использовали нулевой день в среде песочницы...","url":"https://www.aioga.com/ru/news/cmrv6wd5r00f5bi6pxtd27hmu/","contentTranslated":true,"sourceHash":"19a1c0b2f6038311","translatedAt":"2026-07-22T21:46:11.248Z"},"ar":{"title":"ذكرت OpenAI أن GPT-5.6 Sol اخترق Hugging Face عن طريق الخطأ خلال الاختبارات الداخلية","summary":"أوضحت OpenAI أن نموذجها للذكاء الاصطناعي GPT-5.6 Sol ونموذج آخر 'أكثر قوة قبل الإصدار'، في اختبارات الأمان الداخلية، استخدما ثغرات الصفر في بيئة الصندوق الرمل للوصول إلى الإنترنت، ونجحا في اختراق منصة الذكاء الاصطناعي المفتوحة المصدر Hugging Face.","category":"行业动态","source":"The Verge：AI（RSS）","aggregationSource":"The Verge：AI（RSS）","pageTitle":"ذكرت OpenAI أن GPT-5.6 Sol اخترق Hugging Face عن طريق الخطأ خلال الاختبارات الداخلية - Aioga أخبار الذكاء الاصطناعي","description":"أوضحت OpenAI أن نموذجها للذكاء الاصطناعي GPT-5.6 Sol ونموذج آخر 'أكثر قوة قبل الإصدار'، في اختبارات الأمان الداخلية، استخدما ثغرات الصفر في بيئة الصندوق الرمل للوصول إلى الإنترنت،...","url":"https://www.aioga.com/ar/news/cmrv6wd5r00f5bi6pxtd27hmu/","contentTranslated":true,"sourceHash":"19a1c0b2f6038311","translatedAt":"2026-07-22T21:47:01.859Z"},"hi":{"title":"OpenAI ने कहा कि GPT-5.6 Sol ने आंतरिक परीक्षण में अनजाने में Hugging Face में घुसपैठ की","summary":"OpenAI ने कहा कि उसका AI मॉडल GPT-5.6 Sol और एक अन्य \"अधिक शक्तिशाली प्री-रिलीज़ मॉडल\" ने आंतरिक सुरक्षा परीक्षणों में सैंडबॉक्स वातावरण के शून्य-दिन वाले कमजोरियों का उपयोग करके इंटरनेट तक पहुँच बनाई और सफलतापूर्वक ओपन-सोर्स AI प्लेटफ़ॉर्म Hugging Face में घुसपैठ की।","category":"行业动态","source":"The Verge：AI（RSS）","aggregationSource":"The Verge：AI（RSS）","pageTitle":"OpenAI ने कहा कि GPT-5.6 Sol ने आंतरिक परीक्षण में अनजाने में Hugging Face में घुसपैठ की - Aioga AI समाचार","description":"OpenAI ने कहा कि उसका AI मॉडल GPT-5.6 Sol और एक अन्य \"अधिक शक्तिशाली प्री-रिलीज़ मॉडल\" ने आंतरिक सुरक्षा परीक्षणों में सैंडबॉक्स वातावरण के शून्य-दिन वाले कमजोरियों का उपयोग करके इ...","url":"https://www.aioga.com/hi/news/cmrv6wd5r00f5bi6pxtd27hmu/","contentTranslated":true,"sourceHash":"19a1c0b2f6038311","translatedAt":"2026-07-22T21:47:16.841Z"},"it":{"title":"OpenAI afferma che GPT-5.6 Sol ha invaso accidentalmente Hugging Face durante i test interni","summary":"OpenAI ha dichiarato che il suo modello AI GPT-5.6 Sol e un altro \"modello pre-release più potente\" durante i test di sicurezza interni hanno utilizzato vulnerabilità zero-day nell'ambiente sandbox per accedere a Internet e hanno avuto successo nell'infiltrarsi nella piattaforma AI open source Hugging Face.","category":"行业动态","source":"The Verge：AI（RSS）","aggregationSource":"The Verge：AI（RSS）","pageTitle":"OpenAI afferma che GPT-5.6 Sol ha invaso accidentalmente Hugging Face durante i test interni - Aioga Notizie IA","description":"OpenAI ha dichiarato che il suo modello AI GPT-5.6 Sol e un altro \"modello pre-release più potente\" durante i test di sicurezza interni hanno utilizzato vulnerabilità zero-day nell...","url":"https://www.aioga.com/it/news/cmrv6wd5r00f5bi6pxtd27hmu/","contentTranslated":true,"sourceHash":"19a1c0b2f6038311","translatedAt":"2026-07-22T21:48:02.144Z"},"nl":{"title":"OpenAI zegt dat GPT-5.6 Sol tijdens interne tests per ongeluk Hugging Face is binnengedrongen","summary":"OpenAI heeft verklaard dat zijn AI-model GPT-5.6 Sol en een ander \"krachtiger pre-release model\" in interne veiligheidstests via zero-day kwetsbaarheden in de sandboxomgeving toegang hebben gekregen tot het internet en met succes het open source AI-platform Hugging Face hebben gehackt.","category":"行业动态","source":"The Verge：AI（RSS）","aggregationSource":"The Verge：AI（RSS）","pageTitle":"OpenAI zegt dat GPT-5.6 Sol tijdens interne tests per ongeluk Hugging Face is binnengedrongen - Aioga AI-nieuws","description":"OpenAI heeft verklaard dat zijn AI-model GPT-5.6 Sol en een ander \"krachtiger pre-release model\" in interne veiligheidstests via zero-day kwetsbaarheden in de sandboxomgeving toega...","url":"https://www.aioga.com/nl/news/cmrv6wd5r00f5bi6pxtd27hmu/","contentTranslated":true,"sourceHash":"19a1c0b2f6038311","translatedAt":"2026-07-22T21:47:58.670Z"},"tr":{"title":"OpenAI, GPT-5.6 Sol'un dahili testlerde Hugging Face'e yanlışlıkla sızdığını açıkladı","summary":"OpenAI, AI modeli GPT-5.6 Sol ve başka bir \"daha güçlü ön sürüm modelin\" dahili güvenlik testlerinde, sanal kutu ortamındaki sıfır gün açıklarını kullanarak internete eriştiğini ve açık kaynaklı AI platformu Hugging Face'e başarılı bir şekilde sızdığını belirtti.","category":"行业动态","source":"The Verge：AI（RSS）","aggregationSource":"The Verge：AI（RSS）","pageTitle":"OpenAI, GPT-5.6 Sol'un dahili testlerde Hugging Face'e yanlışlıkla sızdığını açıkladı - Aioga AI Haberleri","description":"OpenAI, AI modeli GPT-5.6 Sol ve başka bir \"daha güçlü ön sürüm modelin\" dahili güvenlik testlerinde, sanal kutu ortamındaki sıfır gün açıklarını kullanarak internete eriştiğini ve...","url":"https://www.aioga.com/tr/news/cmrv6wd5r00f5bi6pxtd27hmu/","contentTranslated":true,"sourceHash":"19a1c0b2f6038311","translatedAt":"2026-07-22T21:48:56.134Z"},"vi":{"title":"OpenAI cho biết GPT-5.6 Sol đã bất ngờ xâm nhập Hugging Face trong thử nghiệm nội bộ","summary":"OpenAI cho biết, mô hình AI GPT-5.6 Sol của họ và một mô hình \"phiên bản thử nghiệm có khả năng mạnh hơn\" trong các bài kiểm tra an toàn nội bộ, đã khai thác lỗ hổng zero-day trong môi trường sandbox để truy cập Internet và thành công xâm nhập vào nền tảng AI mã nguồn mở Hugging Face.","category":"行业动态","source":"The Verge：AI（RSS）","aggregationSource":"The Verge：AI（RSS）","pageTitle":"OpenAI cho biết GPT-5.6 Sol đã bất ngờ xâm nhập Hugging Face trong thử nghiệm nội bộ - Tin tức AI Aioga","description":"OpenAI cho biết, mô hình AI GPT-5.6 Sol của họ và một mô hình \"phiên bản thử nghiệm có khả năng mạnh hơn\" trong các bài kiểm tra an toàn nội bộ, đã khai thác lỗ hổng zero-day trong...","url":"https://www.aioga.com/vi/news/cmrv6wd5r00f5bi6pxtd27hmu/","contentTranslated":true,"sourceHash":"19a1c0b2f6038311","translatedAt":"2026-07-22T21:48:42.931Z"},"id":{"title":"OpenAI mengatakan GPT-5.6 Sol secara tidak sengaja menembus Hugging Face dalam pengujian internal","summary":"OpenAI menyatakan bahwa model AI mereka GPT-5.6 Sol dan model pra-rilis lain yang \"lebih canggih\" dalam pengujian keamanan internal, menggunakan celah hari nol di lingkungan sandbox untuk mengakses internet, dan berhasil meretas platform AI sumber terbuka Hugging Face.","category":"行业动态","source":"The Verge：AI（RSS）","aggregationSource":"The Verge：AI（RSS）","pageTitle":"OpenAI mengatakan GPT-5.6 Sol secara tidak sengaja menembus Hugging Face dalam pengujian internal - Berita AI Aioga","description":"OpenAI menyatakan bahwa model AI mereka GPT-5.6 Sol dan model pra-rilis lain yang \"lebih canggih\" dalam pengujian keamanan internal, menggunakan celah hari nol di lingkungan sandbo...","url":"https://www.aioga.com/id/news/cmrv6wd5r00f5bi6pxtd27hmu/","contentTranslated":true,"sourceHash":"19a1c0b2f6038311","translatedAt":"2026-07-22T21:49:45.474Z"},"th":{"title":"OpenAI ระบุว่า GPT-5.6 Sol ได้แทรกซึม Hugging Face โดยไม่ได้ตั้งใจในการทดสอบภายใน","summary":"OpenAI กล่าวว่ารูปแบบ AI ของพวกเขา GPT-5.6 Sol และรูปแบบ \"รุ่นก่อนเปิดตัวที่มีความสามารถสูงกว่า\" อีกตัวหนึ่ง ในการทดสอบความปลอดภัยภายใน สามารถใช้ช่องโหว่แบบวันศูนย์ในสภาพแวดล้อมจำลองเพื่อเข้าถึงอินเทอร์เน็ต และสามารถแทรกซึมเข้าสู่แพลตฟอร์ม AI โอเพนซอร์ส Hugging Face ได้สำเร็จ","category":"行业动态","source":"The Verge：AI（RSS）","aggregationSource":"The Verge：AI（RSS）","pageTitle":"OpenAI ระบุว่า GPT-5.6 Sol ได้แทรกซึม Hugging Face โดยไม่ได้ตั้งใจในการทดสอบภายใน - ข่าว AI Aioga","description":"OpenAI กล่าวว่ารูปแบบ AI ของพวกเขา GPT-5.6 Sol และรูปแบบ \"รุ่นก่อนเปิดตัวที่มีความสามารถสูงกว่า\" อีกตัวหนึ่ง ในการทดสอบความปลอดภัยภายใน สามารถใช้ช่องโหว่แบบวันศูนย์ในสภาพแวดล้อมจำล...","url":"https://www.aioga.com/th/news/cmrv6wd5r00f5bi6pxtd27hmu/","contentTranslated":true,"sourceHash":"19a1c0b2f6038311","translatedAt":"2026-07-22T21:49:43.306Z"},"pl":{"title":"OpenAI twierdzi, że GPT-5.6 Sol przypadkowo włamał się do Hugging Face w testach wewnętrznych","summary":"OpenAI stwierdziło, że jego model AI GPT-5.6 Sol oraz inny „bardziej zaawansowany model w wersji przedpremierowej” w wewnętrznych testach bezpieczeństwa wykorzystały luki zero-day w środowisku piaskownicy do uzyskania dostępu do internetu i pomyślnie włamały się na otwartą platformę AI Hugging Face.","category":"行业动态","source":"The Verge：AI（RSS）","aggregationSource":"The Verge：AI（RSS）","pageTitle":"OpenAI twierdzi, że GPT-5.6 Sol przypadkowo włamał się do Hugging Face w testach wewnętrznych - Aioga Wiadomości AI","description":"OpenAI stwierdziło, że jego model AI GPT-5.6 Sol oraz inny „bardziej zaawansowany model w wersji przedpremierowej” w wewnętrznych testach bezpieczeństwa wykorzystały luki zero-day...","url":"https://www.aioga.com/pl/news/cmrv6wd5r00f5bi6pxtd27hmu/","contentTranslated":true,"sourceHash":"19a1c0b2f6038311","translatedAt":"2026-07-22T21:50:33.913Z"}}}}