Good morning. Have you read the gossipy story about Elon Musk going to Washington in The Atlantic?
Come for the unbelievable exchanges (“Fuck you! Fuck you! Fuck you!” Treasury Secretary Scott Bessent was shouting at Elon Musk in the halls of the West Wing last month…), stay for the commentary on the management style of the Tesla CEO-turned-DOGE…well, doge. (“We probably didn’t need the chain saw. We needed the chisel.”)
I haven’t been served tea this hot since I was in London for our last Brainstorm AI, is all I’m saying—perfect for a Friday read. Have a great weekend. —Andrew Nusca
Fortune Tech is off Monday, May 26 for the Memorial Day holiday. We’ll be back in your inboxes Tuesday, May 27.
Want to send thoughts or suggestions to Fortune Tech? Drop a line here.
Anthropic releases two new AI models
Anthropic—the Chevrolet to OpenAI’s Ford—unveiled on Thursday two new AI models that set new benchmarks in their respective areas of capability.
Anthropic’s new Claude Opus 4 (“our most powerful model yet,” the Dario Amodei-led company says) is a model focused on coding and agent workflows.
Its new Claude Sonnet 4 model meanwhile emphasizes so-called reasoning, which allows it to break down complex tasks into logical steps. (“The model balances performance and efficiency,” Anthropic says.)
Both models are “hybrid” models that balance rapid response and “extended thinking” modes that adjust their processing effort as desired. They both can remember the context of conversations over longer periods of time than their predecessors.
The latest release allows Anthropic to leapfrog—for now, anyway; there’s a new AI model released every other week lately—its pure-play AI peers as it works to justify its latest valuation of $61.5 billion. Anthropic is currently the seventh most valuable private company in the world, behind OpenAI (No. 3) and ahead of xAI (No. 8), according to CB Insights.
The new models are available via Anthropic’s API, Amazon Bedrock, and Google Cloud Vertex AI with pricing in line with the AI models they replace. —AN
Apple will reportedly release smart glasses at the end of 2026
Ever wish Apple abandoned its $3,500 Vision Pro in favor of something simpler that would compete directly with Meta’s popular Ray-Bans? This item’s for you.
The tech giant reportedly aims to release smart glasses at the end of next year.
“Apple will start producing large quantities of prototypes at the end of this year with overseas suppliers,” according to a new Bloomberg report.
The frames would have cameras, microphones, and speakers, just like Meta’s specs. It would also have Siri.
The glasses would be used to make phone calls, play music, offer turn-by-turn directions, and translate language on the fly, according to the report.
“One person with knowledge of the glasses said they will be similar to the Meta product but better made,” it reads. (Ouch.)
Apple is working on a dedicated chip for the product, the report adds, and is exploring other wearables. (AI-powered devices: so hot right now!)
Meta and Apple are approaching the category from opposite ends of the spectrum. While Apple started at the high end and is working its way down to a more accessible product, Meta is doing the reverse, reportedly developing a pricier model that includes a display and, ultimately, augmented reality. —AN
U.S. lawmakers pass budget bill that includes AI law ban
The U.S. House of Representatives on Thursday narrowly passed a budget bill that, if enacted, would ban states from enforcing laws regulating artificial intelligence for a period of 10 years.
As is typical for a budget bill, H.R. 1 covers a lot of ground: taxes, immigration, and other pet topics for the Trump administration. (The House vote approving the bill was largely along party lines; Republicans currently control a majority of the body.)
But it’s the AI language that’s notable for this newsletter.
Today there is no comprehensive federal legislation directly regulating artificial intelligence in the U.S., leaving that duty to the states.
If this bill was signed into law, a 10-year moratorium on state AI laws would represent the first major federal action on the subject—and one of the biggest federal moves on tech policy in a generation.
Its success is hardly a fait accompli.
Democratic lawmakers are almost certain to challenge the AI moratorium portion of the bill in a reconciliation package.
Republicans are no lock, either. Some legislators have expressed concern that they would be overriding desirable laws (e.g. deepfake protections) in the states they represent. —AN
More tech
—Tinder CEO steps down. Faye Iosotaluno is out after less than two years; Match Group CEO Spencer Rascoff temporarily takes over.
—Disney sues YouTube. Alleged breach of contract after the Google-owned video service hired veteran exec Justin Connolly as its media and sports chief.
—FTC won’t block Microsoft acquisition. The agency drops its case against the Activision Blizzard deal, almost two years after it closed.
—R.I.P. Pocket. The read-it-later service, acquired by Mozilla in 2017, calls it a day on July 8.
—Big Banks explore joint stablecoin. Companies co-owned by JPMorgan Chase, Bank of America, Citi, Wells Fargo, and others are reportedly in talks.
—OnlyFans may sell. Owner Fenix is in talks to move the not-quite-all-adult-entertainment service to an investor group…or go public.
—U.S. charges 16 people for DanaBot malware. The Russia-based malware operation infected more than 300,000 machines around the world.
Endstop triggered
