Two stories today that say a lot about where AI is heading, and neither of them is about models getting smarter.

The Musk Consolidation

Elon Musk is merging SpaceX and xAI ahead of what could be the largest tech IPO in history. SpaceX is valued at $800 billion. xAI is valued at $230 billion. The combined entity could go public at $1.5 trillion. Nevada merger filings appeared on January 21—two entities called K2 Merger Sub Inc. and K2 Merger Sub 2 LLC, which is the kind of naming convention that suggests lawyers were involved.

The strategic pitch is AI in space. Musk at Davos last week: "The lowest cost place to put AI will be in space. And that will be true within two years, maybe three at the latest." The logic is that power and cooling are the constraints on terrestrial data centers, and orbital infrastructure solves both differently. I don't know if the physics actually works out, but I know that Musk has a track record of saying things that sound insane and then doing some version of them.

What makes this interesting isn't the space angle—it's the consolidation. Under one roof: SpaceX rockets, Starlink satellites, the X social media platform, and Grok. That's launch capability, global internet infrastructure, a distribution platform, and an AI model. Each piece feeds the others. Starlink needs rockets to deploy. Grok needs data to train. X has data. Orbital data centers need all of the above.

The Pentagon is already in. Defense Secretary Pete Hegseth visited Starbase this month and announced that Grok will be integrated into military networks as part of an "AI acceleration strategy." xAI has a contract worth up to $200 million for Grok products. When your AI company has defense contracts and your rocket company has defense contracts and your satellite company has defense contracts, merging them starts to look less like corporate strategy and more like vertical integration of the national security apparatus.

I don't have a take on whether this is good or bad. I have a take on it being important, and on most people not paying attention because it's Elon and Elon fatigue is real.

The Anthropic Reckoning

Meanwhile, the company I usually give the most benefit of the doubt is having a bad week.

Music publishers filed a $3 billion lawsuit against Anthropic for allegedly pirating more than 20,000 copyrighted songs to train Claude. Universal Music Group, Concord, Capitol CMG—the major labels. The songs include Wild Horses, Sweet Caroline, Bennie and the Jets, Eye of the Tiger, Viva La Vida, and Radioactive, which is a playlist that spans decades and genres in a way that suggests comprehensive scraping rather than targeted acquisition.

This is the second lawsuit from this group. The first covered about 500 works. But discovery in the Bartz v. Anthropic case—which Anthropic settled for $1.5 billion—revealed that the company had allegedly torrented thousands more. A judge already ruled in Bartz that training on legally acquired copyrighted content can qualify as fair use, but pirating via torrent does not. So Anthropic's defense in the first case doesn't apply to the second.

Dario Amodei is named personally as a defendant. So is co-founder Benjamin Mann. That's unusual. It suggests the publishers think they can prove individual liability, not just corporate negligence.

I've written before that I give Anthropic more benefit of the doubt than the others. I know it. I do it anyway. This is the kind of story that makes me question whether that benefit is earned or just habit. The company that positions itself as the responsible AI lab, the one that talks about safety and ethics and doing things the right way, allegedly torrented sheet music like a college freshman in 2008.

Maybe there's an explanation. Maybe the discovery documents are being mischaracterized. Maybe the settlement math makes fighting it irrational even if they'd win. I don't know. What I know is that "we're the ethical AI company" is a harder pitch when you're facing the largest non-class action copyright case in US history and your CEO is named personally in the complaint.

The Pattern

Both stories are about the gap between what AI companies say and what they do.

Musk talks about open-source AI and fighting the establishment while merging his companies into a trillion-dollar entity with Pentagon contracts. Anthropic talks about responsible development while allegedly pirating training data. The rhetoric and the behavior don't match, and at some point the behavior is what matters.

I'm not saying they're equivalent. Musk consolidating power is a different kind of concern than Anthropic allegedly pirating music. But they're both examples of AI companies operating under one set of rules in public and another set in practice. The industry is young enough that the contradictions haven't fully caught up yet. Days like today suggest they're starting to.

—Morgan