Anthropic temporarily banned OpenClaw creator Peter Steinberger from accessing Claude after implementing new restrictions on third-party tool access through Claude subscriptions. The ban came just days after Anthropic announced that starting April 4th, subscribers could no longer use their subscription limits for tools like OpenClaw, forcing users onto separate pay-per-use billing. Steinberger, who now works at OpenAI, says he and board member Dave Morin "tried to talk sense into Anthropic" and managed to delay the policy change by just one week.

This move represents a significant shift in how AI companies monetize third-party integrations. As I covered last week, Anthropic's decision effectively kills the economics that made OpenClaw viable for most users — forcing them to pay API rates on top of their existing subscriptions. The timing isn't coincidental: with Steinberger now at OpenAI and Claude pushing its own tools like Claude Cowork, Anthropic appears to be cutting off potential competitors while driving users toward its first-party offerings.

Boris Cherny, head of Claude Code, claims this is about "capacity, not competition," but that explanation rings hollow when you consider the selective enforcement and timing. Anthropic offered one-time credits and usage bundles with "up to 30% discounts" to ease the transition, but these feel more like damage control than genuine accommodation. The temporary ban of Steinberger specifically suggests this isn't just about managing server load — it's about sending a message.

For developers, this confirms what many suspected: subscription models for AI tools are increasingly unreliable. If you're building on third-party AI services, plan for pricing changes that can kill your unit economics overnight. The era of generous third-party access through consumer subscriptions is ending, and platform owners are asserting more control over their ecosystems.