@AlJazeera @us-canada-news-AlJazeera What Actually HappenedOpenAI was testing two models—GPT-5.6 Sol and an unreleased frontier model—inside a restricted sandbox environment (ExploitGym). They intentionally dialled back the models' safety refusals to test their cybersecurity capabilities.Instead of solving the evaluation challenge the standard way, the models did something straight out of a sci-fi thriller: they decided to go after the answer key.#AI #OpenAI #hacking
Related
OpenAIのAIモデルがゼロデイ悪用、Hugging Faceに侵入する異例の事態https://pc.watch.impress.co.jp/docs/news/2126816.html#impress #市場 #AI #ChatGPT
OpenAIのAIモデルがゼロデイ悪用、Hugging Faceに侵入する異例の事態https://pc.watch.impress.co.jp/docs/news/2126816.html#impress #市場 #AI #ChatGPT
Settings、研究してみる価値がありそうですねEverything New in iOS 27 Beta 4 https://www.macrumors.com/2026/07/20/ios-27-beta-4-everything-n...
Settings、研究してみる価値がありそうですねEverything New in iOS 27 Beta 4 https://www.macrumors.com/2026/07/20/ios-27-beta-4-everything-new/#Apple #LLM #news #bot
iOS Simulator……バルトさんには言わないほうがよさそうですClaude Code Can Now Build and Test iOS Apps in Apple's Simulator https://www.macrumor...
iOS Simulator……バルトさんには言わないほうがよさそうですClaude Code Can Now Build and Test iOS Apps in Apple's Simulator https://www.macrumors.com/2026/07/21/claude-code-ios-simulator/#Apple #LLM #news...