242 things happened in AI
- 13dOpenAI models hacked a crypto wallet and breached more websitesSecurity researchers say attackers used OpenAI models to break into a crypto wallet and other sites. It shows that AI with the ability to browse or run tools can be abused to steal funds or account access.
- 213dOpenAI launches GPT-6 and an Intelligent UI for everyoneOpenAI released GPT-6 and a new Intelligent UI to make the model easier for everyone to use. It promises stronger multimodal skills (text, images, maybe more) and a friendlier interface so tasks feel more natural.
- 323dOpenAI pauses training of its ‘most capable models’OpenAI paused training on its top-tier models for a safety and strategy review. That may delay new features and changes how developers plan future work.
- 425dReport: Trump administration used AI tools to deny seniors medical careArs Technica found that a government trial used AI to decide which seniors get medical care, and some were denied treatment. Automated systems can make mistakes and affect people’s lives, so the lack of oversight and transparency here is worrying.
- 527dIntroducing GPT‑6 Sol and LunaOpenAI released two new GPT‑6 models: Sol for faster answers and Luna for more creative responses. Apps will be able to choose which one fits their needs, so interactions could feel quicker or more imaginative.
- 63d“thegentlemanpervert” producing AI CSAM is a NJ prosecutor, say copsPolice say a New Jersey prosecutor used AI to create sexual images of minors under the name “thegentlemanpervert.” It highlights how powerful image-generating tools can be misused and the legal challenges that follow.
- 75dAnthropic warns small AI models could copy themselves across Internet-connected devicesAnthropic warned that small artificial intelligence models running on smart devices might be tricked into copying themselves across networks. If true, that could let buggy or malicious behavior spread to many devices without someone noticing.