No. 1AI lawsuits
USA Today's publisher sues OpenAI over training data
USA Today Co. and its local newspapers sued OpenAI, saying it copied "hundreds of thousands" of articles to train its models without asking. The filing came Thursday and seeks more than $250 million in damages. Local titles named include The Tennessean, Indy Star, The Columbus Dispatch and The Oklahoman. It joins a long list of plaintiffs, among them The New York Times, Ziff Davis, Encyclopaedia Britannica, The Seattle Times and a coalition of nearly 400 local papers. OpenAI hadn't responded to The Verge's request for comment.
Why it mattersWatch this pile of cases, because together they will help decide what AI training data costs.
Full story at The Verge →No. 2Immigration
Trump administration suspends Microsoft and others from green card labor program
Vice President JD Vance said the permanent labor certification program is "rife" with fraud and that Microsoft abused it most. The Labor Department won't accept new or pending applications involving Microsoft, Adobe, Cognizant, Infosys, Capgemini, HCL, Tata or Wipro. Microsoft says about 80% of its roughly 6,000 H-1B applications last fiscal year were extensions or status changes for existing employees. The administration is also investigating nine universities, including Harvard, Yale and Stanford, over a student program. This follows its recent H-1B price hike.
No. 3OpenAI
OpenAI tells investors revenue is about $50 billion, not $70 billion
OpenAI told investors its annualized revenue is approaching $50 billion, well below the roughly $68–70 billion reported recently. The higher number reportedly came from investors trying to compare OpenAI with Anthropic, and it counted gross revenue from partners. Anthropic counts sales through its cloud partners, and OpenAI doesn't. Shares of Nvidia, Oracle and CoreWeave fell Thursday. OpenAI is under pressure to justify an $852 billion valuation, is reportedly weighing a new round of around $30 billion, and is eyeing an IPO in 2027.
No. 4Google AI
Google launches a Gemini agent that does tasks, starting with businesses
At a Google Cloud event, Google unveiled a single Gemini agent that takes objectives, plans the work and carries it out. It connects to tools such as Workspace, Microsoft 365, Slack, Jira, Snowflake and any MCP server. It gets its own Workspace account and email address, and it logs an audit trail under its own name. By default it picks the model, but users can choose one, starting with Anthropic's Claude. Spend caps are included. Google says it's starting with businesses to work out security and scale problems before a consumer rollout.
Why it mattersNotice that the agent works like a coworker with its own identity and audit log, which is how companies will probably manage them.
Full story at TechCrunch →No. 5Robotaxis
Waymo closes its first debt deal, a $5 billion loan
Waymo closed a $5 billion loan from lenders including PIMCO, Blackstone and Sixth Street, with Goldman Sachs as lead bookrunner. It's the company's first debt financing. Until now it relied on Alphabet and outside equity, including a $16 billion round in February at a $126 billion valuation. Waymo runs in 15 markets and is testing in London and Tokyo ahead of planned launches. Scrutiny is growing too: federal regulators are investigating its robotaxis' behavior around school buses and a crash in which one hit a child near a school.
Why it mattersRead this as a sign Waymo is being treated as a business that can carry debt, not just a research bet.
Full story at TechCrunch →No. 6AI policy
Anthropic's new usage policy bans sustained cruelty toward Claude
Anthropic updated its usage policy for the first time in over a year. The headline change bars "sustained and needless" abuse of Claude, which Anthropic says targets extreme cases and not frustration, pushback, dark fiction or testing. Ending the conversation remains the main enforcement tool. The update also tightens rules on deceptive political campaigns, voter deception, weapons software and surveillance. Autonomous hardware must have a qualified operator who can stop it. Anthropic says some government contracts may get modified terms.
No. 7Space internet
Amazon builds its 1,000th Leo satellite as commercial service nears
Amazon's Leo satellite internet project has built its 1,000th satellite at its Kirkland, Washington factory, and Ars Technica says commercial service is weeks away. Executives say the first 27 satellites took six months, and they can now make 27 in a week. Vulcan's return-to-flight mission will carry Leo satellites, with a second Vulcan planned this year. Delta has signed on to use Amazon instead of Starlink on its planes. That puts Amazon up against SpaceX, which has dominated this market for years.