OpenAI cuts off Musk, Anthropic BEATS Pentagon, Grok could go DARK: AI NEWS

Link to our newsletter: https://bitbiased.ai/

OpenAI is reportedly cutting off model access to Cursor after SpaceX announced a $60 billion acquisition of Anysphere. A federal judge just ruled that the Pentagon’s blacklisting of Anthropic was illegal and baseless. And xAI has warned a court that an environmental dispute over gas turbines could cause Grok to largely cease functioning for hundreds of millions of users.

Three stories that look completely unrelated are really asking the same question: who actually controls whether an AI model gets to keep running?

The first fight centers on Cursor, the AI coding tool built by Anysphere. After SpaceX announced it was acquiring the company in a deal reportedly worth $60 billion, OpenAI moved to cut off Cursor’s access to its models, with the cutoff scheduled for November 12. According to the reporting discussed in this video, OpenAI’s justification involves distrust of Musk-connected companies alongside alleged terms-of-service violations tied to xAI.

But that explanation has two very different layers. A terms-of-service violation is something concrete. Distrust of companies connected to Elon Musk is much broader. And when the decision arrives within days of one of OpenAI’s best-funded competitors acquiring Cursor, the move becomes more than a simple compliance story.

There is also an important limitation: we do not yet have a public, itemized breakdown showing exactly which terms were allegedly violated and when. That makes this potentially both a policy enforcement decision and a strategic competitive move.

Then there is Anthropic.

Earlier in 2026, the Department of Defense pushed Anthropic to remove two restrictions governing military use of Claude: its prohibition on autonomous weapons selecting targets without human involvement and its prohibition on mass domestic surveillance of Americans. Anthropic refused.

The Pentagon subsequently designated Anthropic under a federal statute that legal experts described as having never previously been used against a domestic American company and as normally being associated with foreign sabotage threats. Anthropic sued in March.

Now a federal judge has ruled that the Pentagon’s blacklisting of Anthropic was illegal and baseless.

That could reopen the path toward military contracts potentially worth billions, but it does not mean those contracts automatically return. The Pentagon can still appeal, and existing arrangements involving competitors including OpenAI and xAI remain unaffected. What changed is Anthropic’s legal position after months of fighting one of the most powerful customers in the world over how Claude could be used.

And then the third story moves from model access and military policy to something much more physical: electricity.

xAI is facing a Clean Air Act lawsuit involving gas turbines at a Mississippi power site that the suit alleges were not properly permitted. Those turbines help support the compute infrastructure behind Grok.

In a federal court filing, xAI argued that shutting down that power source could cause Grok to largely cease functioning for hundreds of millions of users.

That is an enormous claim, but it needs context. It comes from xAI itself during active litigation in which the company is attempting to avoid an injunction. It should therefore be understood as a legal argument rather than an independently verified assessment that Grok is actually on the verge of going offline.

Put all three stories together and the bigger pattern becomes clear.

OpenAI can control access to the models another company depends on. Anthropic just fought over how much leverage the Pentagon can exercise over an AI provider that refuses certain uses of its technology. And xAI is arguing that regulators’ control over physical power infrastructure could ultimately determine whether Grok stays available.

Same lever. Three different hands trying to hold it.

The most important AI battles may increasingly have less to do with benchmark scores and model capabilities and more to do with access agreements, government contracts, courts, infrastructure, electricity, and who ultimately controls the switch.

And none of these stories is finished yet. The Cursor cutoff has not taken effect. The Pentagon can still appeal the Anthropic ruling. And xAI’s turbine dispute remains in litigation.

The models themselves may not have changed this week. The power structures determining whether people can use them did.

CHAPTERS

00:00 Who Actually Controls the AI Switch?
01:47 OpenAI Cuts Cursor Loose
03:49 The Court Just Overturned the Pentagon Ban
06:02 xAI Says Grok Could Go Dark
08:00 What Actually Connects These Three Stories
09:01 The Limitations Nobody Is Putting in the Thumbnail
10:11 The Verdict

#openai #anthropic #xai #grok #ai