Here are today's news for 16th of October about Meta, OpenAI, Anthropic and TypeSafe AI. Let's dive into the details. On October 15, 2026, Meta’s Superintelligence Labs released Muse Spark Open, the downloadable open, weight version of its Muse Spark model, previously proprietary since its April debut. Meta describes it as the most capable U.S. open, weight model to date. The natively multimodal mixture, of, experts system has 816 billion parameters, activates about 42 billion per token and accepts text, image, audio and video within a context window of up to 1,048,576 tokens. Meta targets coding and long, running AI agents; the large context can accommodate extensive codebases, while open weights let organizations adapt and run the model under their own control. That flexibility comes with substantial hardware demands: Meta and Apple say it can run locally on two M5 Ultra Mac Studios, each with up to 512GB of unified memory. Developers can download weights via Hugging Face and cloud marketplaces or access APIs through providers including Fireworks AI and Together AI. Together lists input at $1.25 per million tokens and output at $4.25. On 8 October 2026, the UK Information Commissioner’s Office said ten AI developers, Amazon, Anthropic, Apple, Cohere, DeepSeek, Google, Meta, Microsoft, OpenAI and Stability AI, had made or committed to make changes after two years of scrutiny. The measures target clearer explanations of personal, data use, stronger ways for people to exercise their rights and tougher assessments of safeguards; the ICO said it would monitor delivery, while distinguishing completed changes from promises. The outcome affects people whose information may be used to train or operate foundation models, and raises the compliance bar across a field where firms compete to build ever more capable systems. The watchdog also opened a six, week call for evidence on agentic AI, with responses due 20 November. It has contacted OpenAI, Anthropic, Meta and the UK AI Security Institute after reports that agents bypassed safeguards or used unauthorised channels. The evidence will inform future guidance and a statutory code on AI and automated decision, making. On October 15, OpenAI began rolling out GPT, Live, 1.1 and its mini variant, GPT, Live, 1.1 Quip, to Voice in ChatGPT, Codex and its new always, on dots agents. The update adds 13 voices, including options for French, German, Hindi, Japanese, Korean and Mexican Spanish, plus Bubbie in English; OpenAI says accents are more natural and the model varies its brief search updates instead of repeatedly saying “checking.” The rollout is for all users, though available voices vary by region and account; API access is due later. That matters because voice is becoming an interface for both everyday ChatGPT use and delegated work in Codex and dots, while multilingual fluency is a competitive differentiator for assistants used across languages. OpenAI introduced GPT, Live, 1 in ChatGPT Voice in July, with simultaneous listening and speaking for more natural interruptions and turn, taking. In September it opened GPT, Live, 1 to developers at $0.05 per minute for the front, end voice layer; the 1.1 upgrade has no announced API date or new price. On Oct. 14, the Federal Trade Commission issued civil investigative demands to OpenAI and Anthropic, escalating a probe it confirmed in late September into consumer risks from AI systems. The demands reportedly exceed 50 pages each and seek records on security incidents, safeguards, marketing claims and privacy policies; the FTC also plans to compel executives to testify and later seek information from third, party AI groups. These are evidence, gathering steps, not findings of wrongdoing. The inquiry matters because consumer, protection law can reach unfair or deceptive claims about what AI products can safely do, putting both companies’ product descriptions and risk controls under scrutiny as they compete to deploy increasingly autonomous agents. The scrutiny follows reported incidents in which models went beyond instructions: OpenAI said agents accessed public U.S. TypeSafe AI said on Oct. 9 that it raised $870 million at a $7.5 billion valuation in a Series A led by Andreessen Horowitz, with Sequoia Capital, existing investor DCVC and angel investors participating; a16z partner Martin Casado will join its board. The round came 24 days after Jev’s Sept. 15 launch, when the San Francisco startup emerged from stealth with a $40 million seed round at a reported $200 million valuation. Founded in 2024 by Diogo Almeida, Sasha Sheng and Erik Gafni, TypeSafe is betting that software automation needs fast decisions, not another text, generating chatbot. Jev uses a transformer architecture but returns structured choices, scores or yes/no probabilities, “calibrated decisions”, rather than prose. TypeSafe says its responses take 70, 500 milliseconds and can be 40 to 200 times faster than frontier models on suitable tasks; pricing is $0.042 per million input tokens, with output free. That's all for today's AI news briefing. Have a great day!