Anthropic says the U.S. government ordered it to suspend access to Fable 5 and Mythos 5 over national security concerns. The company disputes the reasoning behind the directive, arguing the alleged jailbreak is limited and that similar capabilities exist in competing AI models. The move could become a major moment in the debate over government oversight of advanced AI.
Anthropic has launched Claude Fable 5 for public use while restricting access to Claude Mythos 5 over concerns tied to cybersecurity, biology research, and advanced AI misuse. The company claims Mythos 5 possesses some of the strongest offensive cyber capabilities ever seen in a commercial AI model.
OpenAI says artificial intelligence may soon automate a significant portion of its own research process. In a new post from Sam Altman and Jakub Pachocki, the company outlines its “third phase,” discusses personal AGI for everyone, and warns about concentrated AI power while continuing to push toward increasingly capable systems.
Google is pitching Gemma 4 12B as a multimodal AI model powerful enough for agentic workflows while still being lightweight enough to run locally on consumer laptops with 16GB of memory. The company’s encoder-free design could make this one of its more interesting AI releases yet.
Microsoft says AI is not replacing humanity but extending human intelligence, though the company also admits current systems still struggle with reasoning and hallucinations.
Google DeepMind is launching an AI for the Planet accelerator in Asia Pacific, but the initiative also raises uncomfortable questions about Silicon Valley influence and dependency on American tech companies.
OpenAI says ChatGPT is becoming better at recognizing dangerous intent over time, including subtle warning signs involving self harm or violence across multiple chats.
Zyphra is betting big on AMD with 15 megawatts of Instinct MI355X GPU capacity as AI companies keep searching for alternatives to NVIDIA.
You can finally buy AI insurance, but what does that actually mean and will it actually pay out when things go wrong?
ChatGPT’s goblin obsession wasn’t a joke. OpenAI says training rewards pushed the AI to overuse weird creature metaphors.
United Imaging Intelligence has open sourced a medical video AI model along with a massive dataset and benchmark, showing how smaller specialized models can outperform general purpose AI in clinical tasks.
A newly released open source framework called SentiAvatar aims to make AI avatars behave more like real people during conversation. The project includes a motion dataset with 21,000 clips, a foundation model trained on hundreds of hours of motion data, and a system that synchronizes speech, facial expressions, and body gestures in real time. If the technology works as advertised, it could help developers build more believable NPCs, AI assistants, and virtual characters.