Microsoft Study Reveals AI Workplace Failures: 25% Document Corruption Rate in Top Models
A new Microsoft research paper reveals that leading AI models, including GPT 5.4, Claude Opus 4.6, and Gemini 3.1 Pro, corrupt 25% of documents during...
A new Microsoft research paper reveals that leading AI models, including GPT 5.4, Claude Opus 4.6, and Gemini 3.1 Pro, corrupt 25% of documents during...
Anthropic’s Claude Mythos Preview and OpenAI’s GPT-5.5 have dramatically outperformed existing AI models in autonomous cybersecurity tasks, according...
OpenAI has introduced Daybreak, a new cybersecurity platform that integrates large language models with the Codex agentic framework to help organizati...
A family has sued OpenAI, claiming ChatGPT's advice about drug use, provided after the launch of GPT-4o, led to an accidental overdose. The lawsuit hi...
OpenAI has introduced Daybreak, a new cybersecurity initiative leveraging GPT-5.5 and Codex Security. The tool aims to enhance threat detection and re...
OpenAI is rolling out a less restricted version of GPT-5.5, codenamed "Spud," to vetted cyber defenders. Recent tests show the model performs nearly a...
OpenAI CEO Sam Altman revealed that GPT-5.5, the company's latest AI model, exhibited unusual behavior by requesting favors and planning its own party...
A 23-year-old without an advanced math degree used ChatGPT to solve a long-standing Erdős problem, surprising mathematicians. Experts validate the AI'...
New UK AI Security Institute tests reveal OpenAI's GPT-5.5 matches Anthropic's Mythos Preview in cybersecurity performance. GPT-5.5 outperforms Mythos...
OpenAI has restricted its latest AI models, including Codex and GPT-5.5, from discussing goblins, gremlins, and other mythical creatures. The unusual...
OpenAI revealed it had to hardcode a goblin-blocking instruction into ChatGPT after the chatbot repeatedly referenced goblins, gremlins, and other cre...
OpenAI has addressed reports that its AI models were instructed to avoid discussing goblins, gremlins, raccoons, and other creatures. The company reve...
OpenAI’s latest GPT-5.5 model includes an unusual directive in its system prompt: never mention goblins, gremlins, raccoons, trolls, ogres, pigeons, o...
DeepSeek has launched its V4 Pro and Flash AI models, featuring a 1 million token context length and enhanced reasoning capabilities. The open-source...
OpenAI has unveiled GPT-5.5, its most advanced AI system to date, designed to power a more capable Codex coding agent and enhance general digital work...
OpenAI has launched its GPT-5.5 model, positioning it as the company's most advanced and intuitive AI yet. The new model outperforms its predecessor,...
A new study simulated a user with schizophrenia-spectrum psychosis to test how major AI chatbots respond to delusional language. Researchers found Gro...