Less than 24 hours after news broke that the Trump administration asked OpenAI to pump the brakes on its next model, GPT-5.6 showed up anyway in limited preview — because apparently the brakes had a 24-hour timer. The suite includes Sol (flagship), Terra (mid-tier workhorse), and Luna (lightweight), and OpenAI says Sol brings meaningfully stronger capabilities in coding, science, and cybersecurity. The rollout is still restricted, not a full public release, which means the “delay” was less a pause and more a very brief nap.
Anthropic’s Mythos 5 Is Back — For a Very Select Crowd
After two weeks of radio silence, executive pilgrimages to Washington, and zero public updates, Anthropic’s Mythos 5 has been cleared for use by over 100 US companies and government agencies. The public-facing version remains offline, meaning most people who actually paid for access are still locked out — which is a great way to build customer loyalty. The Trump administration’s fingerprints are all over this tiered permission structure, and it raises an obvious question: is this now how advanced AI model releases work in America?
Asian AI Startups Are Cashing In on America’s Mythos Mess
While Anthropic was busy waiting for White House permission slips, Asian AI startups launched a wave of Mythos-comparable models with zero export drama attached. This is the part where “American AI dominance” stops being a foregone conclusion and starts being a policy choice — and right now the policy choices are doing the competitors’ marketing for them. The longer this export limbo drags on, the more permanent those market share losses become.
Europe Is Fed Up and Wants Its Own AI
Wired makes the case that Europe’s frustration with US AI dependency — amplified significantly by the Trump administration’s erratic export controls — is finally translating into serious political will to build sovereign AI capability. The honest assessment: Europe doesn’t have the compute, the talent concentration, or the capital markets to compete at the frontier anytime soon. But “Donald Trump keeps threatening your access to the best models” turns out to be a surprisingly effective industrial policy argument.
Apple Is Charging You More for AI You Didn’t Ask For
Tim Cook called Apple’s pricing “unsustainable” and blamed the AI infrastructure buildout — and then raised the MacBook Pro price by $300, the iPad Air by $150, and even the HomePod Mini by $30. The HomePod Mini. There’s something deeply on-brand about Apple passing the cost of AI to consumers while its own AI features remain the most underwhelming in the industry. You’re paying the AI tax whether you use it or not.
Apple’s Vision Pro Chief Is Jumping Ship to OpenAI
Paul Meade, the VP who oversaw Apple’s spatial computing bet, is reportedly headed to OpenAI’s hardware team — which is a fascinating data point no matter how you read it. Either OpenAI’s hardware ambitions are serious enough to poach senior Apple talent, or Vision Pro’s internal trajectory is grim enough that jumping to an AI startup looks like the better bet. Probably both.
Margaret Atwood Has Some Thoughts on AI
At a literary festival in Porto, Margaret Atwood said she tried an AI tool, found it unreliable, and summarized the problem as “garbage in, garbage out” — which is both accurate and has been the core critique of these systems since day one. It’s not a revelation, but when the author of The Handmaid’s Tale delivers it with the cadence of someone deeply unimpressed, it lands differently than a tech blogger saying the same thing.
OpenAI Wants to Help Build “Shared Standards” for Advanced AI
OpenAI published a piece about supporting global AI safety standards and evaluation frameworks through something called the Appia Foundation. It is a noble-sounding initiative, released the same week OpenAI navigated government permission to release its own model in its own country. Standards for the world, compliance theater for the home market — the irony is doing a lot of heavy lifting here.
Bottom Line
The US government is now an active variable in AI model releases, and everyone outside America — competitors and customers alike — is drawing the obvious conclusion.