原文 · 未翻译
The AI industry's platform trap is starting to look a lot like Microsoft's
Anthropic is throttling its new Mythos model for certain tasks while building apps that directly compete with its largest customers. Customers, partners, and investors are pushing back.
Anthropic released Claude Fable 5 this week, a version of its Mythos model, but quietly dialed back performance when customers used it for tasks related to building their own AI software or hardware. After public backlash, the company walked things back slightly, saying it would at least notify customers when a weaker model is being served.
According to The Information, some investors and industry watchers see their fears confirmed. "It's only a matter of time before only the model creators have access to the most powerful models," Martin Casado, general partner at Andreessen Horowitz, wrote on X back in April. Everyone else gets access to smaller, distilled versions, he said. Casado is also an investor and board member at coding startup Cursor, which is both an Anthropic customer and a competitor to Claude Code.
Officially, the throttling is meant to stop foreign adversaries or rival AI labs from using Anthropic's models to improve their own tech. But according to The Information, developers worry that even basic features for AI-powered apps could get restricted. The suspected motive: Anthropic wants to keep the best tech for its own competing products.
Customers turned competitors are getting louder
The model throttling is just one piece of a bigger pattern in the tangled relationships between model providers and customers who suddenly become rivals. According to The Information, Anthropic asked companies like Figma and Canva to act as partners weeks before launching its AI design tool Claude Design. Right before release, Anthropic expanded the feature set so much that the product competed directly with those partners' offerings. Figma pulled out of the partnership, and Anthropic's Chief Product Officer Mike Krieger left Figma's board.
Figma CEO Dylan Field said at a Sequoia Capital event that Anthropic was "not consistently candid in their communications." A year earlier, Anthropic had already launched Claude Code, a coding app that threatened customers like Cursor. Claude Code eventually surpassed both Cursor and Microsoft's GitHub Copilot in revenue. "If there's just one winner in the AI model race, they can push around and stomp their customers," said Quinn Slack, CEO of coding startup Amp and a longtime Anthropic customer.
Behind Anthropic's aggressive moves is an unprecedented revenue surge. According to The Information, annualized revenue grew fivefold in just five months to nearly $50 billion, overtaking OpenAI, which also offers a competing product to many of its own customers with Codex. Combined, Anthropic and OpenAI are growing faster than the next 32 large AI startups put together. Both companies are deepening their moats through subsidiaries like OpenAI's DeployCo.