All sections
News
507 stories · page 33 of 57
InfraSaudi HUMAIN is now selling compute on AMD chips, and the next step is 250 megawattsOn 31 August AMD, Cisco and HUMAIN announced that AMD Instinct MI355X systems are live in Saudi Arabia and serving customers. The next phase is a plan: up to 250 megawatts on the new MI400 generation from 2027. Their joint venture's target remains up to 1 gigawatt by 2030.FrontierDeepSeek released the weights of its vision model under MIT and this time showed the numbersOn 31 August DeepSeek uploaded DeepSeek-V4-Flash-Vision-Exp to Hugging Face under the MIT licence. Ten days earlier the model was API-only, without a single image benchmark. Now the table is out, comparing the new model with their older one and with Opus 4.8.ControlOpenAI backed California's bill on children and AI and praised it for protecting responsible use of ChatGPT's memoryOn 31 August OpenAI urged Governor Newsom to sign SB 1119 - a bill on the safety of minors in AI services. The list of requirements is serious. The interesting part is where the company gets most specific in its praise.DesignRunway showed a model that draws the interface frame by frame, with no code underneathOn 31 August Runway introduced Solaris, the first model in a family it calls Interface World Models. Instead of an app with code underneath, every frame of the interface is generated the moment you touch it. For now it is research with early access, not a product.SecurityTwo PaperCut flaws entered the actively exploited catalogue, and they chain togetherOn 31 August CISA added CVE-2026-82078 and CVE-2026-81578 in PaperCut NG and MF to the KEV catalogue. The two chain into code execution. The vendor said on 27 August that it was investigating active exploitation, and advises closing the web interface to the internet immediately.BuildersOpenAI is pulling its models from Cursor on 12 November because it does not trust the new ownerOn 28 August OpenAI notified SpaceX that it is ending the contract under which its models run in Cursor. It is giving the maximum notice the contract allows and withholding future models, Astra included. The reason is stated plainly: its experience with Elon Musk's companies.ControlClaude found fixes for all ten kinds of bad behaviour Anthropic gave it to treatOn 28 August Anthropic published a report in which Claude trains other models on its own against ten categories of alignment failure. Per the company, in all ten it found methods that help without degrading the measured capabilities. The comparison with humans the company itself calls indirect.ScienceAnthropic previewed a standard that lets an AI agent run a microscope, a liquid-handling robot and a robotic armOn 27 August Anthropic opened a research preview of the Model Hardware Standard - a shared driver through which an agent discovers the instruments in a lab and gives them commands. Integration that, by the company's account, takes weeks drops to hours. The open-source release comes later, after safety evaluations.
507 stories