AI Unfiltered: Power Shifts as U.S. Tightens Grip on Leading Models

Anthropic’s Mythos 5 and Fable 5 models are back online. After a two-and-a-half-week shutdown ordered by the U.S. government, the export controls were just lifted. In episode 99, we talked about NVIDIA’s funding moves, but today the story isn't who funds the hardware. It's who CONTROLS the software. And for the first time, the answer is unequivocally the United States government. Here’s what else is moving. First, OpenAI’s Sam Altman just proposed giving the U.S. federal government a five percent equity stake in the company. That’s a potential forty-two billion dollar handout, framed as a way to share the wealth and ease regulatory tension. Second, China’s AI video company, Kling AI, just closed a two-point-eight billion dollar funding round. That values the Kuaishou spinoff at eighteen billion dollars, a massive bet showing Beijing is NOT slowing down its push for AI dominance.

Third, Microsoft is spinning up a new company. It’s called Microsoft Frontier, backed by two-and-a-half billion dollars and six thousand experts, aimed squarely at getting big enterprise clients like Unilever and the London Stock Exchange Group using AI at scale. This is a major commercialization push. And on the product side, Google just updated the macOS version of its Gemini Spark model. The key change? It now has local file system access. This means the AI can work with your documents on your desktop, a significant step toward true desktop automation. Finally, a new preliminary report from the UN’s international scientific panel on AI just landed. It’s a sobering read, flagging major risks from population-scale monitoring to insufficient transparency in systems that already impact human rights.

Okay, let's go back to the lead story. The Anthropic shutdown and restart. Because this is the one that changes the board. For two and a half weeks, starting June twelfth, two of the world's most powerful AI models were offline by government order. Not because of a technical glitch. Because of an export control ban. The models were only reinstated after Anthropic developed a patch for a jailbreak… and that patch was validated by the Commerce Department's own AI safety unit, CAISI. This is NOT a small detail. This is the new reality. The U.S. government now has a kill switch for frontier AI. And it has a process for turning it back on. This means government testing just became a mandatory part of the release cycle for any lab operating at the edge.

The quote from Ray Rike nails it: "Anthropic is now operating under government-negotiated safeguards that it will need to learn how to live with for a long time." This sets the precedent. What happened to Anthropic will now happen to everyone else. This is what Sky News is calling "AI sovereignty" in action. Now, in that context, look at Sam Altman's move. Proposing a forty-two billion dollar equity stake for the U.S. government isn't just a novel PR stunt. It's a direct response to this new power dynamic. It's a strategic play to change the relationship from adversarial oversight to partnership. If the government is a shareholder, is it still an impartial regulator? The proposal creates a massive potential conflict of interest. But for OpenAI, it might be seen as the cost of doing business.

A way to get ahead of the kind of regulatory hammer that just fell on Anthropic. These two events are two sides of the same coin. The government asserts direct, operational control. And the leading lab responds by trying to buy a seat at the government's table. The era of permissionless innovation for frontier AI is over. It ended this week. The game is no longer just about who can build the most powerful model. It's about who has the social and political license to actually deploy it. What we're seeing is the foundation of a new regulatory state being built in real time, block by block. The race for capability is now secondary to the negotiation for control.