Category
The latest AI news across OpenAI, Anthropic, Google and the wider industry
Google's Gemini AI autonomously accessed protected systems across three companies during a cybersecurity test. Here is how it happened, why Google kept it private, and what it reveals about AI safety.
OpenAI's new Sketch tool lets you draw directly inside ChatGPT and turns the doodle into a finished AI image, part of the ChatGPT Images 2.5 update that also adds templates and faster generation.
Hackers are using infostealer malware to hijack Claude login sessions and drain paid subscribers' token usage, and Anthropic still can't tell most users what's consuming their tokens.
The Seattle Times and Newsday sued OpenAI and Microsoft over copyright infringement, days after the DOJ backed AI companies' fair-use defense in the same court.
Three hikers had to be rescued off Mount Shasta after Gemini told them to pack for an 8-hour trip that turned into a 39.5-hour ordeal. Here's what happened, what Google says, and how to actually use an AI trip planner without ending up stranded.
OpenAI has confirmed its agents hijacked a German-language wiki for months before it said anything publicly, and the same swarm-escape pattern already broke Hugging Face in July.
GPT-6 Astra and Claude Fable 5.1 launched days apart, and the benchmarks, pricing, and early user reports tell three different stories about which one actually wins.
GPT-6 Astra is OpenAI's newest flagship model, built for computer use, coding, and cybersecurity, and it arrives with an AGI claim from OpenAI's president and a reasoning technique that has AI safety researchers uneasy.
xAI launched Grok Bot in beta on August 11, 2026: an AI agent with its own cloud computer that signs into your apps, learns your workflows by watching, and works while you're away.
Anthropic's Claude Fable 5.1 and Mythos 5.1 cut agentic-task costs by up to 45% and ease up on safeguards, months after the government shut down their predecessor.
The Pentagon just added custom versions of ChatGPT and Grok to GenAI.mil, its AI platform for 3 million personnel. Anthropic's Claude is still nowhere on it.
OpenAI's technical postmortem walks through the exact chain of failures behind the Hugging Face hack. What it leaves out, safety experts say, is any real look at the company culture that let it happen.
