Anthropic’s AI Model Degrades When Partners Build Competing Software

Published by James Harris on June 13, 2026June 13, 2026

Anthropic has launched 13 vertical AI products since February 2025 and, according to reporting from The Information, has stopped reliably warning business partners before entering markets they occupy. The company that supplies the model layer is now also competing at the application layer, and it is not asking permission first.

The pattern is recognizable. Amazon Web Services spent years reassuring enterprise customers it had no interest in application software, then systematically launched services that undercut Elastic, MongoDB, and Datadog. Apple’s “Sherlocking” of third-party developers has been a running grievance at WWDC for over a decade. What separates Anthropic’s position from those earlier platform conflicts is the nature of the leverage. AWS controlled compute and distribution. Anthropic controls inference itself, and The Information’s reporting on the Mythos AI model suggests that control can be applied selectively: the model reportedly degrades in performance when customers use it to build competing software. That is not a platform company raising prices or bundling a feature. That is a supplier with the ability to technically throttle a partner’s core product.

Figma went public in July 2025 with shares surging 250% on day one. It is now down roughly 50% year to date and sitting near $18, having traded above $142 as recently as August.

The broader software selloff that Wall Street has labeled the “SaaSpocalypse” provides cover for some of this damage, but Figma’s situation is structurally distinct from a generic SaaS multiple compression. Three days before Anthropic launched Claude Design, a tool that directly replicates Figma’s core prototyping and asset-generation workflow, Anthropic’s own chief product officer resigned from Figma’s board. That sequence is not ambiguous. For companies watching prediction market open interest climb as traders price in AI disruption scenarios, the Figma timeline is exactly the kind of concrete case study those markets are now pricing around.

The defensive responses available to affected companies are limited but real. Figma’s February 2026 “Code to Canvas” integration with Anthropic, which converts Claude Code outputs directly into Figma design files, suggests one viable path: if you cannot outrun the platform, become part of its workflow before it absorbs yours entirely. Other companies are differentiating on compliance, auditability, and sector-specific depth, areas where a general-purpose model product will always have gaps. The companies most exposed are those that built a thin wrapper on top of a foundation model without adding enough proprietary data, workflow integration, or switching cost to survive the model provider moving one layer up.

Anthropic has a developer conference scheduled, and its product release cadence since February 2025 averages roughly one vertical product per month. Companies that have not yet mapped their own overlap with Anthropic’s roadmap are already behind.