Apple bets on Google, Trump and Sanders agree on AI ownership, and Anthropic warns it may lose control of its own models
Three stories from this week that should not make sense but do: Apple ships Siri on a Google model, a socialist and a populist agree on nationalising AI, and the company that built Claude is publicly worried it might not be able to stop it.
This week in AI moved fast enough that the news from Monday already feels like last quarter. Three stories in particular are worth sitting with, not because of the headlines but because of what they reveal about where this industry is actually going.
Apple hands Siri to Google and calls it a strategy
On Monday, June 8, Apple kicks off WWDC 2026 at 10 a.m. PT. It is Tim Cook's final keynote as CEO before he hands the role to hardware chief John Ternus in September. That alone is a milestone. But the product underneath is what matters.
Apple has rebuilt Siri from scratch, and the engine running it is Google Gemini. Not a custom Apple model. Not an in-house large language model. A licensed 1.2-trillion-parameter Gemini model that Apple is paying Google roughly one billion dollars a year to use. The deal was confirmed in a joint statement in January 2026, and it makes Siri the largest commercial deployment of Gemini outside of Google's own products.
What this actually means: Apple is not building a frontier model. While OpenAI built GPT, Google built Gemini, and Anthropic built Claude, Apple signed a cheque and kept its engineering focused on private compute and hardware integration. It is either the most contrarian strategic bet in tech right now, or the slowest way to become irrelevant. There is not much middle ground.
There is a multi-model layer on top of this. Apple is building an Extensions system that lets users switch between ChatGPT, Gemini, and Claude to handle Apple Intelligence features, each with a distinct voice so you know which model responded. Gemini is the default. OpenAI's exclusivity inside the iPhone, which began with iOS 18, is over.
The strangest political alliance of 2026
On Friday, June 6, President Trump told reporters the US government should take direct equity stakes in AI companies like OpenAI, Anthropic, and xAI. Days earlier, Senator Bernie Sanders published an op-ed in the New York Times calling for a 50% public ownership stake in the same companies. Two politicians who agree on almost nothing are now describing the same policy outcome in different language.
Trump called it a partnership. Sanders called it public ownership. The underlying logic is identical: the infrastructure being built right now is too important to leave entirely in private hands.
The practical likelihood of government equity stakes in OpenAI is low in the near term. But the fact that this conversation is happening at all, and that it has support from both edges of the political spectrum, tells you something about how public trust in the AI industry is shifting.
Anthropic's public warning about its own models
The most unsettling story of the week did not involve a product launch or a political statement. Anthropic issued a rare public warning this week, stating that its own models may soon be too powerful to control. The company used the phrase "brake pedal" in describing what it believes the industry needs before the next generation of systems ships.
This is not performative safety theater. Anthropic has been the most consistent voice on frontier risk since its founding, and this kind of public statement carries weight precisely because it comes from the people who are actively building the thing they are warning about. They are not commentators. They are the engineers.
Worth noting: Project Glasswing, Anthropic's internal frontier safety program, reportedly found over 23,000 vulnerabilities in its most advanced model within the first month of restricted deployment. They are publishing the number. That is unusual. Most companies in this position would bury it. Whatever you think of Anthropic's commercial interests, that level of transparency about capability risk is not nothing.
Taken together, these three stories point at the same underlying shift. AI has stopped being a feature and started being infrastructure. The policy questions, the ownership questions, and the safety questions are all expressions of the same recognition: the systems being built now are not neutral tools. They have stakeholders, they have risks, and they have consequences that go well past the companies that built them.