Anthropic starts selling its text AI to startups

The OpenAI defectors are officially open for business.

Anthropic has quietly begun supplying its text-generating AI models to a select group of commercial startups, moving beyond purely academic research and closed testing.

Legal tech gets the first commercial key

The first paying customer stepping out into the open is Robin AI, a legal tech platform that recently raised over $13 million from backers like Plural and Google’s Black Founders Fund.

Robin AI fine-tuned an Anthropic model on legal documents to help draft and negotiate commercial contracts. CEO Richard Robinson said Anthropic’s heavy focus on AI safety matches their "lawyer-in-the-loop" design, which keeps humans around to catch hallucinations.

Anthropic’s tech is also running inside Quora’s experimental Poe app, but that integration remains unmonetized.

Neither Robin nor Anthropic will say whether the model in use is Claude or a custom variant. In fact, Anthropic refused to share how many startups are currently in the loop or how many models it eventually plans to commercialize.

When the compute bills come due

Anthropic was founded in 2021 by Dario Amodei, OpenAI’s former VP of research, alongside several colleagues who split over OpenAI’s aggressive commercial direction. They structured Anthropic as a public benefit corporation focused heavily on alignment.

Their headline breakthrough was "Constitutional AI" — training models like Claude to follow principled rules instead of relying solely on messy human feedback. Late last year, Claude’s private beta showed genuine promise, though it still occasionally tripped over factual errors and toxic prompts.

Safety research doesn't pay for GPU clusters.

Anthropic has taken in hundreds of millions in capital, including a $580 million round from a group featuring FTX’s Sam Bankman-Fried, followed by Google’s recent $300 million investment for a 10% stake and preferred cloud partnership. Investors want to see revenue, making a commercial API push inevitable.

Why it matters

OpenAI finally has a heavyweight challenger in the enterprise API market.

For enterprise developers, having Anthropic enter the arena provides a serious alternative, especially for regulated workflows where Anthropic's alignment-first reputation could be a genuine selling point. That said, with access strictly gated and model details kept vague, most teams will have to wait before testing it against GPT-3.5 or GPT-4.

Anthropic is now entering the same commercial rat race it was founded to escape. Going head-to-head with OpenAI, Cohere, and AI21 Labs will test whether a public benefit corporation can balance rigorous safety principles with the ruthless pace of enterprise SaaS sales. Partnering with vertical apps that keep a human in the loop is a clever starting line, but the real test begins when scale kicks in.