The handshake era is over

The handshake era of AI is over.

For the past two years, the big AI companies operated on a polite fiction: that the industry was a collaborative ecosystem where models, tools, and distribution channels could be freely mixed and matched. Anthropic licensed Claude to Amazon. Microsoft built its flagship products on OpenAI's models. Third-party developers piggybacked on subscription APIs. Everyone played nice because nobody had enough of the stack to go it alone.

That fiction collapsed in 48 hours this week.

VentureBeat reported that starting today, Anthropic is cutting off third-party tools like OpenClaw from using Claude subscriptions. If you built a product that relied on users authenticating with their Claude accounts, that's done. Users now need to buy usage bundles or get their own API key. Anthropic cited "outsized strain" on compute systems, offered a one-time credit equal to subscribers' monthly plan cost, and shut the door. The timing isn't subtle: OpenClaw's creator Peter Steinberger recently left for OpenAI.

Meanwhile, TechCrunch reported that Microsoft shipped three foundation models built entirely in-house: MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2. These aren't research experiments. MAI-Transcribe-1 beats OpenAI's Whisper on all 25 languages tested and Google's Gemini Flash on 22. MAI-Voice-1 generates a minute of audio in one second with custom voice cloning. MAI-Image-2 debuted as a top-three model on Arena.ai. All three are available now through Microsoft Foundry with aggressive enterprise pricing. Microsoft spent billions on OpenAI and is now systematically building alternatives.

And then there's the strangest move of the three. TechCrunch also reported that OpenAI acquired TBPN, the daily tech talk show hosted by John Coogan and Jordi Hays, for low hundreds of millions. A company that makes AI models now owns a media property that hosts interviews with Zuckerberg, Nadella, and every other tech CEO. The show reports to OpenAI's chief political operative Chris Lehane. There are "contractual editorial independence provisions," which is the kind of phrase that sounds reassuring until you think about it for thirty seconds.

What the pattern means

Each move is different in the specifics but identical in the logic: own more of the stack.

Anthropic is pulling its model behind its own paywall, forcing the ecosystem onto terms it controls. Microsoft is reducing its dependence on a single supplier by becoming one itself. OpenAI is buying narrative distribution because, ahead of an IPO, controlling how your story gets told matters as much as the technology.

This is classic vertical integration, and it happens in every technology cycle when the initial period of open experimentation gives way to platform consolidation. In the browser era it was Netscape then Internet Explorer. In mobile it was the wild west of app stores before Apple and Google locked them down. The pattern is predictable: first you build the technology, then you build the walls.

For anyone building products on AI, the practical consequence is simple. The APIs, subscription models, and integration points you rely on today are policy decisions by companies that are actively rethinking their boundaries. Anthropic can change who gets to use Claude subscriptions overnight. Microsoft can build a competitive model and offer it at cost. OpenAI can buy the media outlet that covers your industry.

The question for builders isn't whether vertical integration is coming. It arrived this week. The question is how much of your product depends on the continued goodwill of a company that just decided goodwill is a luxury it can no longer afford.

---