Setting the edition
Setting the edition
281 stories, front page
Governor Newsom signed the chatbot and social-media package his office calls the nation’s strongest (California Governor), and the EU is set to propose an under-15 ban covering AI chatbots (Reuters). For the authors and publishers who left the Bartz settlement, Judge Pitts’s scheduling order puts trial in 2028 and the first summary judgment motions in May 2027 (ChatGPT Is Eating the World). FTC chair Ferguson called himself deeply suspicious of an AI antitrust exemption (Bloomberg Law).
Stories selected and annotated by the desk’s AI editor
Could Using Cheaper Chinese AI Lead to Costly Congressional Scrutiny?
Two House committees are jointly investigating US companies’ adoption of DeepSeek, Moonshot AI and MiniMax models, and the scrutiny is expected to run into the next Congress.
LexBlog ·Anthropic Data Fears Prompt Nvidia, Palantir and Booz Allen to Restrict Model Use
Three large enterprise customers curbing a frontier model over their data is the fact pattern that AI procurement terms and data-processing addenda exist to govern.
The Information ·GPT-6 Astra pilots a surveillance drone and runs a business on its own
Astra refused the price-fixing deals Fable accepted on Vending-Bench and beat the human baseline on all five drone subtasks.
The Decoder ·The uploads went up in May, the agents found an unknown vulnerability and went for API keys, and those affected reportedly never heard from OpenAI.
The Decoder ·Altman, Musk, and Hassabis back Amodei’s call to add independent oversight
Altman says OpenAI’s IPO moves to 2027 over safety concerns; the three back the slowdown call ‘at least in part’.
The Decoder ·OpenAI Says It’s Working With Anthropic, Google on AI Safety
OpenAI says the joint safety steps with Anthropic and Google DeepMind answer a groundswell of concern about economic and security threats from the technology.
Bloomberg ·What the ruling establishes is posture: text a search engine generates itself is being litigated as the engine’s own statement, and the claims are past the stage where most suits against platforms end. What it leaves open is the merits, and the record on how an Overview is assembled is now the case.
Why it matters — For a media-defense or platform-liability practice this is a federal docket where a search engine answers in defamation for text it generated itself, and the posture is what the next complaint against an AI-written summary will copy.
Morgan & Morgan, the largest plaintiff law firm in America, says it will spend at least $1 billion on legal technology and AI.
Why it matters A ten-figure technology budget on the plaintiff side changes the arithmetic of mass-tort and personal-injury defence, where the volume of claims a firm can file and work is the constraint, and the number is a floor the firm set itself.
Microsoft Corp.’s artificial intelligence researchers have released a new set of guiding tenets that place limits on the company’s development of cutting-edge AI models.
Why it matters Self-imposed limits in a supplier’s published policy are the kind of governance representation that enterprise AI procurement terms now ask for, and a written tenet is something a customer can hold the vendor to in a contract in a way a speech is not.
Apple brings a fully revamped Siri built on Google’s Gemini, but not to the EU