Anthropic faces mounting pressure from customers, partners, and investors over deliberate restrictions on its new Mythos model paired with the company's launch of competing applications. The move mirrors a familiar pattern that plagued Microsoft during its antitrust battles in the 1990s and early 2000s: leveraging platform control to favor proprietary products while limiting what others can build.
Anthropic has implemented throttling on certain Mythos capabilities, constraining what customers can do with the model while simultaneously releasing Claude-branded applications that directly compete with those same customers' businesses. This creates an obvious conflict. Customers funding Anthropic's development face a handicapped model while watching the company extract value by building products in their space.
The tension reflects a structural problem in the AI platform economy. Companies like Anthropic occupy an uncomfortable position. They sell raw model access to enterprises and developers, but they also want to build high-margin consumer and business applications. The incentive to restrict model capabilities becomes clear when doing so protects the company's own product lines.
Microsoft faced similar accusations decades ago. The software giant bundled Internet Explorer into Windows while claiming technical necessity, then restricted what competitors could achieve on the platform. Regulators investigated whether Microsoft abused its monopoly position. The company eventually paid settlements and faced lasting reputational damage.
The parallel isn't perfect. Anthropic doesn't have Microsoft's market dominance, and AI models aren't operating systems. But the playbook is recognizable. Platform owners have structural advantages: they see how customers use the technology, they control the defaults, they can adjust capabilities arbitrarily, and they can allocate resources to their own apps before satisfying customer needs.
This creates what investors and partners are calling a "platform trap." Customers hesitate to build deeply on a platform when the owner might throttle capabilities or launch competing products. Investors worry about backing startups that