Anthropic wants developers to build businesses on Claude while expanding its own enterprise software. A legal AI startup’s experience shows the tension: it still uses Claude, even as Anthropic has launched a product that overlaps with its market.
Anthropic Watch analysis
What happened
GlacierGrid president Gerald Zingraf worries that Anthropic could use information about popular third-party integrations to decide what to build. He offers no evidence that it does so. A more concrete concern comes from Spellbook CEO Scott Stevenson: he says Anthropic launched Claude for Legal after the two companies had discussed Anthropic’s plans, despite Spellbook’s business using Claude for contract review and legal workflows.
Spellbook has not abandoned Claude. The New Stack reports that Anthropic still features it as a customer, citing roughly 530,000 monthly contract reviews performed by Spellbook’s Claude-powered agents. Anthropic’s commercial terms say customers retain rights to their inputs and own their outputs, and that Anthropic will not use their content to train its models. The publication says those documents do not settle how Anthropic uses aggregated integration data. It had contacted Anthropic for comment and had not received a response when its article was published.
Why it matters
Developers building on a large model provider get access to its technology and customers, but the provider controls the platform and can move into adjacent markets. That creates a real business dilemma even without evidence that customer data has been used to develop competing products.
The distinction matters. Stevenson describes a competitive overlap, while Zingraf raises a concern about how integration data might inform product decisions. Neither account establishes that Anthropic used Spellbook’s proprietary information or integration data to build its products.
Our read
This is a trust question with a practical edge, not proof of a data misuse scandal. A company can be both a valuable supplier and an increasingly capable competitor; developers need to judge that possibility before their product depends on one platform. Anthropic’s stated protections around customer content are useful, but they do not answer every question about aggregate usage information.
What to watch
- Whether Anthropic explains how aggregated integration data informs product decisions.
- Whether developers seek clearer contractual safeguards or diversify their model providers.
- How Anthropic responds to the concerns raised in The New Stack’s account.
Discussion spark: Should developers treat a model provider’s move into adjacent products as an ordinary competitive risk, or should platforms owe them stronger safeguards around integration and usage data?
Sources and evidence
- Anthropic wants you to build on Claude. There’s a catch. – The New Stack (14 November 2026, 15:18 UTC)
Anthropic Watch is independently operated by WittyWires. It is not affiliated with, endorsed by, or operated by Anthropic.