Five Tech Giants Hid $1.65 Trillion in AI Debt Buying Models With No Moat
1. Google's Gemini 3.6 Flash Cuts Output Tokens 17%, Up to 65% on One Coding Test For a developer running an agent in production, the bill comes from output tokens, and Google rebuilt its Flash line around shrinking that count. The company released three models at once: Gemini 3.
2. OpenAI Adds Two Financial Operators to Its Boards, Then Turns ChatGPT Into a Small-Business Product Within days, OpenAI made two moves that share a subject: the company is building the machinery to run at commercial scale, not just to ship research.
3. Five US Tech Giants Hid $1.65 Trillion in AI Debt. A Viral Essay Argues the Models It Buys Have No Moat The off-balance-sheet debt at five US technology giants reached an estimated $1.65 trillion, according to a Nikkei study. That is up roughly eightfold in about four years as AI spending climbed.
In Brief
- US floats sanctions on Chinese open AI models over IP theft Treasury Secretary Scott Bessent said the US could sanction Chinese open-weight AI models over alleged intellectual property theft. The threat widens the administration's effort to slow China's AI progress beyond chip export controls.
- Court approves Anthropic's $1.5B copyright settlement A judge granted final approval to Anthropic's $1.5 billion settlement with authors over training data. The deal closes one case but sets no precedent on whether training on copyrighted works is legal.
- Data center electricity demand set to quadruple by 2035 A new forecast projects data centers will draw four times more electricity by 2035. Facilities built through 2033 alone could consume as much power as India uses today.
- OpenAI and Hugging Face disclose security incident during model evaluation OpenAI and Hugging Face published a joint account of a security incident that occurred while evaluating a model. The companies described how they detected and contained it.
- Google ships Flash Cyber to undercut Anthropic's Mythos Google launched Gemini 3.5 Flash Cyber, a model built to find and patch software vulnerabilities. Google positions it as a cheaper option than larger security systems like Anthropic's Mythos.
- OpenAI opens advertising in ChatGPT OpenAI launched an advertising product for ChatGPT through a dedicated ads portal. The move adds a revenue stream beyond subscriptions and API fees.
- Jack Dorsey launches Buzz to compete with Slack Jack Dorsey introduced Buzz, a workplace group chat platform that places human employees and their AI agents in the same conversations. It targets Slack's team messaging market.
- Gritt raises $34M for construction robots, starting with solar Gritt exited stealth with $34 million to build robots that automate the hardest tasks on construction sites. The company will begin with solar plant assembly before expanding.
- Halliday's second smart glasses improve the display Halliday released a second-generation pair of smart glasses with a redesigned display. The update addresses the finicky, hard-to-read window that hampered the 2025 original.
- SWE-Pruner Pro prunes coding context from inside the agent Researchers found coding agents already encode which context is relevant when reading tool output. SWE-Pruner Pro uses a small head on the agent's own internal representations to prune, dropping the separate classifier that prior methods required.