AI Model Changes Spark Outrage Among Developers
A recent decision by AI company Anthropic to restrict how third-party tools can use its services has caused a significant backlash. The move has angered developers who relied on Anthropic’s popular Claude models for their own creations, leading to accusations of unfair practices and a potential shift in the AI development landscape.
Anthropic Restricts Third-Party Access
Anthropic, known for its advanced AI models like Claude, has changed its subscription policies. Previously, users could subscribe to plans like Claude Pro for $20 per month or higher tiers, which offered a generous amount of AI “tokens” – the basic units of text AI models process. These plans were, in effect, subsidized by Anthropic, allowing users to run AI models more extensively than the subscription cost might suggest.
This created an opportunity for third-party applications, such as the open-source tool OpenClaw. Developers used these affordable Anthropic subscriptions to power their own AI agents and complex projects. For example, one user reported spending $200 in a single week on Claude Sonnet 4.6 tokens while running OpenClaw, demonstrating how quickly usage could exceed the subscription cost if not subsidized.
The “Cut Off” and Accusations of Copying
On April 4th, Anthropic began blocking third-party tools, like OpenClaw, from using their subscriptions to access the AI models. While Anthropic stated this was due to “unoptimized” use by these tools, which cost the company more to run compared to their own first-party applications like Claude Code, many developers felt blindsided.
The core of the argument from critics, like Peter Steinberger, the creator of OpenClaw, is that Anthropic is engaging in a “copy then close” strategy. They allege that Anthropic closely studied popular open-source tools, incorporated similar features into their own official products (like Claude Code), and then cut off the third-party developers who innovated these features in the first place.
This feels particularly unfair to those who helped build the ecosystem around Claude. Many users and developers had become vocal advocates for Claude, praising its capabilities, especially its perceived “emotional intelligence” and conversational style, which they felt surpassed competitors like OpenAI’s GPT models.
Technical Explanations and Developer Grievances
Anthropic explained that its own tools, Claude Code and Claude Co-work, were engineered to improve “prompt cache hit rates.” This means they were designed to remember and reuse previous computations for common queries, reducing the overall processing power needed. They claimed that third-party tools like OpenClaw either bypassed these optimizations or underutilized them, making them significantly more expensive for Anthropic to support.
However, critics argue that these technical reasons are a smokescreen for a desire to control their user base. By forcing all third-party applications to pay for API access on a per-token basis, rather than offering subsidized flat-rate subscriptions, Anthropic effectively makes it much more expensive to build on their platform outside of their own controlled products.
Some users have also reported that even mentioning OpenClaw within the system prompt could prevent Claude Code from running commands, suggesting Anthropic is actively trying to discourage the use of such tools. This is in contrast to OpenAI, which has explicitly encouraged the use of third-party tools with its models.
Why This Matters
This situation highlights a critical tension in the AI industry: the balance between innovation, open development, and commercial interests. Developers who relied on affordable access to powerful AI models are now facing increased costs or the need to switch to alternative platforms.
The perception that Anthropic copied innovations from the open-source community before cutting off access could damage trust. For startups and individual developers, this creates uncertainty about building future applications on Anthropic’s platform, as they fear similar actions could occur again. It raises questions about whether AI companies will prioritize fostering an open developer community or focus solely on controlling their own product ecosystems.
The Broader Context and Future Outlook
The controversy comes at a time when Anthropic, like many AI companies, is focused on scaling its business, with potential IPO plans in the future. The company argues that these changes are necessary for business survival, preventing massive financial losses from heavily subsidized token usage. Without these changes, they claim, subscription prices for all users would have to increase, or service quality would decline.
Despite the business rationale, the backlash is significant. The move has led to a negative sentiment, with many former advocates now speaking out against Anthropic. The situation is further complicated by recent security incidents, including a leak of Claude’s internal workings and a large-scale DMCA takedown that affected thousands of GitHub repositories.
Interestingly, Peter Steinberger, the creator of OpenClaw, has since joined OpenAI. This move could allow him to bring his expertise in creating AI with a unique “soul” or personality to OpenAI’s GPT models. Meanwhile, the open-source community continues to innovate, with OpenClaw recently adding a “dreaming” feature, mirroring a capability Anthropic was reportedly developing.
Ultimately, Anthropic’s decision has forced a difficult conversation about fairness, transparency, and the future of AI development. While the company may have acted within its rights, the way it handled the transition has alienated a core group of its most passionate users and developers.
Source: Claude just changed overnight (YouTube)
