16 October 2026
Meta releases open-weight Muse Spark Open, built to run on Macs
Meta led the day with a new open-weight model that Meta says runs on Apple hardware. ChatGPT gained a Chrome tabs feature and reports say it targets ads using your prompts. Regulators and AI labs also made moves on oversight and data.
Models
Meta launches open-weight AI model that runs on Macs (opens the source in a new tab)
Meta and Apple have teamed up on Muse Spark Open, which will run locally on Apple silicon.
Why it matters: If you want AI that keeps your data on your own machines, this is a new option to test.
TechCrunch · 15 Oct
Meta and Apple make Muse AI models run better on Macs (opens the source in a new tab)
Meta calls Muse Spark Open the most capable US open-weight model to date. It says the model can run on two Apple M5 Ultra Mac Studios with the MLX framework.
Why it matters: Running it takes high-end hardware, so check the cost before you plan around it.
The Verge · 15 Oct
Muse Spark Open: Meta's open-weight multimodal model on Hugging Face (opens the source in a new tab)
Hugging Face introduces Muse Spark Open, an open-weight multimodal mixture-of-experts model from Meta.
Why it matters: Open weights mean your developer can download and test it without a vendor contract.
Hugging Face · 15 Oct
LFM2.5-24B-A2B built for fast on-device agents (opens the source in a new tab)
Liquid AI released a model named LFM2.5-24B-A2B, built for fast agents that run on a device.
Why it matters: On-device models can cut cloud costs and keep customer data local, so ask your developer if they fit.
liquid.ai via Hacker News · 14 Oct
Mistral introduces Timestral for forecasting (opens the source in a new tab)
Mistral AI says Timestral is a probabilistic forecasting model in four sizes, from 5M to 209M parameters. It says it runs efficiently on CPUs and works zero-shot.
Why it matters: If you forecast sales or demand, a small model that runs on ordinary hardware may be worth a test.
Mistral AI · 15 Oct
Tools and products
ChatGPT can now see your Chrome tabs (opens the source in a new tab)
OpenAI says its ChatGPT Chrome extension can now view your open tabs when you tag Chrome in the web or desktop app. The Atlas browser is gone.
Why it matters: Check what is open in your browser before you tag it, especially client or private pages.
The Verge · 15 Oct
HighLevel AI Studio adds Git Sync with GitHub (opens the source in a new tab)
You can connect GitHub to an AI Studio project and keep code in sync both ways. Edits made in tools like Claude Code, Cursor or Copilot come back into the project.
Why it matters: If you build in HighLevel, your code is now backed up and can be edited in other tools.
HighLevel · 15 Oct
HighLevel AI Studio Payments gets checkout upgrades (opens the source in a new tab)
You can choose which checkout fields buyers fill in and use a two-step checkout to capture leads before payment. A full order summary shows after payment.
Why it matters: A two-step checkout lets you follow up with buyers who leave before paying.
HighLevel
Security and safety
FTC reportedly probes Anthropic and OpenAI (opens the source in a new tab)
A report says the FTC is looking at Anthropic and OpenAI. Ars Technica says it may fine AI firms on par with companies that delay data breach reports.
Why it matters: Rules on AI companies are tightening, so keep a record of which vendors hold your data.
Ars Technica · 15 Oct
ChatGPT reportedly targets ads using prompts, phone number and email (opens the source in a new tab)
404 Media reports that ChatGPT targets ads based on your prompts, phone number and email.
Why it matters: Check your privacy settings and avoid pasting client details into accounts that may be used for ads.
404media.co via Hacker News · 15 Oct
Anthropic gives METR employee-level access (opens the source in a new tab)
METR will publish its own assessments of Anthropic's risks, incidents and practices. Anthropic has given up the right to block them.
Why it matters: If you rely on Claude, independent reports may help you judge its risks.
Enterprise
Sullivan & Cromwell builds Agreement Analyzer with OpenAI (opens the source in a new tab)
With OpenAI, the law firm developed an AI tool that identifies issues in a draft agreement.
Why it matters: Contract review is a common use of AI, so ask your own lawyer how they use it.
OpenAI · 15 Oct
Anthropic revenue reportedly reaches $16.8B in third quarter (opens the source in a new tab)
The Financial Times reports Anthropic's third quarter revenue jumped to $16.8B.
Why it matters: Strong revenue suggests Claude is widely paid for, which makes it a safer long-term choice.
ft.com via Hacker News · 14 Oct
Meta survey says most small businesses find AI boosts productivity (opens the source in a new tab)
Meta research across nine APAC markets says more than three quarters of small business owners see productivity gains. Most say they need more guidance.
Why it matters: This is Meta's own research, but it shows many peers want training and clear steps.
Meta · 15 Oct
Research
14,000 calls suggest humans still beat AI in call centres (opens the source in a new tab)
A study from boldagents.co analysed 14,000 calls. It says humans still outperform AI.
Why it matters: If you plan to replace phone staff with AI, test it on real calls first. The source is a vendor.
boldagents.co via Hacker News · 15 Oct
Google DeepMind uses Antigravity to search for antimicrobial peptides (opens the source in a new tab)
Google DeepMind describes using Google Antigravity for Science to speed up the search for new antimicrobial peptides.
Why it matters: It shows where Google's agent tools are heading, though it is mainly for science.
Google DeepMind · 15 Oct
Indic pretraining: 2.8 trillion tokens and an open data pipeline (opens the source in a new tab)
Hugging Face published work on Indic language pretraining using 2.8 trillion tokens, with an open data generation pipeline.
Why it matters: Better Indic language models may help if you serve customers in those languages.
Hugging Face · 16 Oct
Also worth knowing
YouTube Music adds AI labels for music (opens the source in a new tab)
YouTube Music is rolling out labels for AI-generated music. They will come to YouTube soon.
Why it matters: If you use AI music in content, expect it to be labelled.
Ars Technica · 15 Oct
OpenAI shows Study Mode for teachers in Brazilian Portuguese (opens the source in a new tab)
OpenAI released AI for Educators in Brazilian Portuguese and shows how Study Mode supports step-by-step learning.
Why it matters: If you run training or coaching, Study Mode is a useful pattern for guided learning.
OpenAI · 15 Oct