OpenAI eyes Washington as a shareholder
The day's biggest story: OpenAI is reportedly proposing to hand the Trump administration a stake of roughly 5% in the company. According to a Financial Times report picked up across The Verge, Bloomberg, CNN and Reuters-syndicated outlets, the move would make the U.S. government a part-owner of one of the world's most valuable AI firms. It's a striking blurring of the line between private industry and the state at the exact moment AI has become a matter of national strategy — and it sets the tone for a day defined by the tightening bond between AI labs and Washington.
Anthropic's détente — and its crackdown
Anthropic sits on both sides of that government relationship. Its two most powerful models, Claude Fable 5 and Mythos 5, are coming back online after a standoff with the U.S. government: the company says the Department of Commerce has lifted the export controls that had sidelined them. At the same time, Anthropic is tightening the screws elsewhere. Per the FT (via Techmeme), it's moving to shut the indirect back-door routes Chinese firms have used to reach Claude, stepping up enforcement against companies accessing the models through intermediaries. Rounding out a busy stretch, Anthropic also shipped Claude Sonnet 5 on June 30 — a midrange model SiliconAngle reports can handle complex jobs more independently than prior Sonnet versions, chasing Opus-level capability at a lower price point.
Meta's rough patch
If OpenAI and Anthropic had a strong day, Meta had a candid one. CEO Mark Zuckerberg used a Thursday town hall to tell staff the company's AI push has moved more slowly than he'd hoped, per Reuters. Superintelligence chief Alexandr Wang struck a more upbeat note, telling employees that Meta's upcoming model — codenamed Watermelon — is closing the gap and now matches OpenAI's GPT-5.5, according to Business Insider. Meta also quietly launched Pocket, a social app for creating and sharing small, interactive AI 'gizmos,' as reported by The Verge and Engadget. The mixed signals — slower-than-hoped internally, catching-up externally, shipping consumer experiments — capture a company still finding its footing.
Microsoft goes all-in on adoption
While others race on model capability, Microsoft is betting on getting businesses to actually use AI. CNBC reports the company is standing up a new operation called Frontier Company, backed by a $2.5 billion commitment and 6,000 employees, tasked with helping other businesses understand and put AI to work. It's a wager that the bottleneck now is adoption, not raw model horsepower.
The skeptics speak up
Not everyone is sold on how the boom is being run. Palantir CEO Alex Karp blasted the industry's token-based pricing — the standard way leading labs charge customers — as 'effing insane,' taking direct aim at OpenAI and Anthropic. On a deeper note, prominent researcher Yann LeCun told the BBC that today's AI isn't actually intelligent, arguing current systems fall short of genuine understanding — and that he's building what comes next.
Redefining intelligence in the classroom
That question — what real intelligence is — is playing out in schools too. As AI-powered tools move into education, Yahoo Finance reports a wave of new products alongside a deeper debate about what learning is actually for. It's a fitting coda to a day that saw AI push further into government, business and the classroom all at once.