Here are today's news for 25th of October about Microsoft, Nvidia, Meta and 22 other companies. Let's dive into the details. On Oct. 23, a coalition of more than 20 technology companies, including Nvidia, Microsoft, Meta, IBM, Hugging Face and Mistral, urged U.S. policymakers to avoid broad restrictions on open, weight AI models. The letter warned that “premature restrictions” could “stifle competition or drive innovation overseas”; OpenAI and Anthropic were absent. Open, weight systems let users download model parameters, modify them and run models on their own infrastructure, unlike proprietary services whose weights remain controlled by providers. That distinction affects startups, researchers and organizations seeking to customize AI or avoid dependence on a handful of vendors, while raising concerns about misuse and the distillation of capabilities from closed models. The coalition argued those concerns should be addressed through “targeted legal and commercial frameworks,” not sweeping limits, and called for wider access to computing and shared datasets, tools and evaluation frameworks. On October 8, the UK Information Commissioner’s Office (ICO) reported that 10 major foundation, model developers had made or committed to make changes after two years of scrutiny: Amazon, Anthropic, Apple, Cohere, DeepSeek, Google, Meta, Microsoft, OpenAI and Stability AI. The regulator said the changes include clearer explanations of how personal data is used, stronger routes for people to exercise data rights, and more rigorous assessments of safeguards; it is monitoring delivery, and the report sets no common deadline. The commitments affect people whose information may appear in training data or model outputs, while putting pressure on competing developers to show that large, scale data collection complies with UK GDPR. The ICO’s supervision examined lawful basis, transparency, access and objection rights, and sensitive data. It warned that developers must address whether models can contain personal information and how special, category data, such as health or religious beliefs, is used. Anthropic opened a research preview of Claude Duet on Oct. 21, 2026, with the launch reported Oct. 24. The proactive agent runs in its own cloud virtual machine and browser, so tasks continue after users close the app or switch off their devices. It can use more than 6,000 connectors, including Gmail, Google Calendar, Slack, QuickBooks, HubSpot and DocuSign, to prepare work such as email drafts, invoice reminders and morning briefings. The waitlist is initially limited to U.S. subscribers on Claude Pro and Max; Duet draws on the user’s existing Claude memory, skills and connected apps. Anthropic’s preview positions Claude as an agent that initiates and follows up on work, rather than only responding to prompts, intensifying competition in AI assistants and workplace automation. The company says users control app access and can disconnect services; tasks needing human judgment can be sent for approval. OpenAI made its Decisions API generally available on Oct. 22, 2026, powered exclusively by GPT, 6 Luna. The endpoint returns bounded answers, probabilities, selections with confidence scores, or numeric ratings, instead of generating a full response, for tasks such as classification, routing and deciding what an agent should do next. OpenAI says it can make decisions up to 10 times faster than GPT, 6 Luna through the Responses API. Developers can send text, images and structured JSON; each request now supports 400 questions, double the public, beta limit of 200, and the release adds GPT, Live integration for voice applications. Pricing is $0.10 per million uncached input tokens and $0.01 per million cached input tokens, with no output, token charge. US and EU regional processing is available, as is zero data retention for eligible customers. The API began public beta on Oct. At GTC Berlin on October 21, 2026, Nvidia introduced DSX POD, a reference design for complete AI factories built around its Vera Rubin platform. Rather than shipping individual servers for on, site integration, partners will assemble and validate the systems in modular containers before delivery. Nvidia says that at a prepared site, deployment could shrink from six months to as little as 14 days; deliveries from partners including Dell Technologies, HPE, Lenovo, Schneider Electric and Vertiv are expected to begin in early 2027. The announcement targets a major constraint on AI expansion: getting power, cooling, networking and compute installed quickly enough to meet demand. Faster commissioning could help cloud providers and enterprises bring capacity online sooner, while shifting more integration work to equipment partners. DSX POD builds on Nvidia’s broader effort to sell complete “AI factory” infrastructure, not just GPUs, as it prepares Vera Rubin systems for the next generation of large, scale AI workloads. That's all for today's AI news briefing. Have a great day!