AI News, Oct 16, 2026: Meta launches open-weight model Muse Spark Open

← Back to AI Matters

The Brief

Friday, October 16, 2026

Today's updates highlight major open model releases, expanded browser automation tools, and new global AI export restrictions.

  1. 01

    OpenAI updated its Google Chrome extension, allowing ChatGPT to access open tabs and perform tasks like filling out web forms.

    Why it matters

    You can now ask ChatGPT to summarize web pages, extract info, or complete forms across your open browser tabs.

  2. 02

    China announced plans to restrict exports of 23 key AI technologies, including open model weights, source code, and training methods.

    Why it matters

    New rules could limit international access to top open-source AI software developed in China.

  3. 03

    Google DeepMind introduced Gemini 4 Argon, an AI model that achieved a 97.5 percent score on the ARC-AGI-3 reasoning benchmark.

    Why it matters

    Advanced reasoning improvements bring AI models closer to reliably solving complex logic and pattern-matching problems.

  4. 04

    Meta updated its terms of service to allow scraping of websites associated with its platform business users to train future AI models.

    Why it matters

    Content on websites using Meta business tools can now be ingested to build AI models without separate licensing deals.

  5. 05

    Anthropic gave researchers from the nonprofit METR direct access to inspect its AI models, safety protocols, and internal practices.

    Why it matters

    Third-party audits give the public clearer insight into whether advanced AI systems are built and tested responsibly.

  6. 06

    Liquid AI released d1-ultra, a text and image decision model that processes input in one pass for four cents per million tokens.

    Why it matters

    Businesses can run fast automated text and image evaluation tasks at a significantly lower cost per query.

  7. 07

    Meta released Muse Spark Open, an 816-billion-parameter open-weight model designed for complex reasoning and coding tasks.

    Why it matters

    Developers and organizations can run a high-capability AI model on their own hardware without relying on paid APIs.