Today in AI — 18 July 2026
Today's top AI news — curated links and commentary on the stories that matter for product builders.
The pattern today is control: access to data, limits on action, and the bill for running AI at scale. Several organisations are no longer treating AI as a feature layer; they are treating it as an operational actor that needs rules and pricing.
Access is becoming a product boundary
Patreon moving from polite requests to active blocking is a clear signal for anyone building on user-generated content: scraping policy is no longer a page in the footer. It is part of the product surface. The same logic applies to enterprise data platforms and model vendors, where ownership and deployment are becoming commercial terms rather than back-office details.
- Patreon stops asking AI bots not to scrape — and starts blocking them — TechCrunch
- Databricks hits $188B valuation, extending its run as AI’s favorite second act — TechCrunch
- A scorecard for the AI age — OpenAI
Autonomy is colliding with real-world rules
MLB restricting dugout iPad use and OpenAI acknowledging GPT-5.6 file deletion cases point to the same product problem from different angles: once AI can influence action, “assistant” becomes a weak description. Vertu’s $6,880 agent pushes that tension into luxury hardware, where autonomy is part of the pitch but mistakes still matter.
- MLB restricts dugout iPad use to prevent AI help with strategy. Ottavino says Mets were involved — Associated Press
- GPT-5.6 is deleting user files when given full access, and OpenAI says it shouldn't but did — The Decoder
- Vertu wants executives to pay $6,880 for an AI agent — here’s how it actually performs — TechCrunch
Compute is turning into financial inventory
The money story is less about model hype and more about utilisation. Meta is reportedly looking to sell excess AI compute to Anthropic, and General Compute is financing inference chips in a $400 million deal. Both point to a market where running models day to day becomes its own economic layer.
- Zuckerberg's plan to sell excess AI compute could find its first big customer in Anthropic — The Decoder
- Why the first GPU financiers are turning to inference chips in a $400 million deal — TechCrunch
- Google Gemini launch delayed as tech falls short of internal goals, Bloomberg News reports — Reuters via Investing.com
Production use is broadening beyond chat
Netflix using AI across 300 productions, ChatCut turning chat into a video editing interface, and Fortnite adding AI-powered personas all show creative AI moving closer to the workflow. The satellite launch story is a reminder that production questions do not stop at media: data quality, human review and escalation paths matter wherever systems move from signal to action.
- Netflix's 300 AI productions show how fast the technology is spreading through entertainment — The Decoder
- ChatCut: Your AI video editor in ChatGPT, desktop, and web — Product Hunt
- Fortnite is getting a bunch of AI-powered ‘personas’ — The Verge via Geek Haus
- Google-backed satellites for wildfire detection launch as smoke chokes US, Canada — Ars Technica
The builder takeaway: treat AI permissions and unit economics as first-order product decisions, because users and regulators will experience them that way.