Anthropic races to contain the Claude Code source leak

Anthropic PBC is urgently dealing with the fallout from the unintentional publication of the internal source code behind Claude Code, the company's AI assistant and one of its key sources of revenue, Bloomberg Tech reports.

What Claude Code is and why it matters

Claude Code is Anthropic's AI assistant for developers. It works as an agent that can write, edit and debug code directly in the terminal or an integrated development environment, drawing on models from the Claude family. In recent years, such tools, agentic coding assistants, have grown from niche experiments into one of the main sources of commercial growth for large AI labs, because developers and companies are willing to pay for subscriptions that noticeably speed up writing and maintaining code. That is why Bloomberg Tech stresses that Claude Code is not just one of Anthropic's products but a key source of the company's revenue.

The market for agentic coding assistants has grown considerably in recent years: besides Anthropic's own products, other major players offer similar tools for automating code writing and review. In such a competitive environment, the technical implementation becomes a significant competitive advantage that companies try not to disclose publicly. That includes how the agent loop is organized, which system instructions govern the model's behavior, and how work with tools and the file system is structured.

• Company: Anthropic PBC
• Product: Claude Code, an AI assistant for development
• Incident: unintentional publication of internal source code
• Product significance: one of the company's key sources of revenue
• Source: Bloomberg Tech

Why is a source code leak serious?

For a company that earns money from a closed product, a leak of its internal source code potentially means several risks at once. First, competitors may gain insight into the technical implementation, such as the agent's architecture, system prompts and tool orchestration logic, which usually remain trade secrets. Second, publishing internal code may expose undocumented vulnerabilities that attackers could exploit before the company manages to close them. Third, for enterprise customers who pay for Claude Code precisely because they count on the closed, secure nature of Anthropic's product, the leak itself is a reputational blow, regardless of what exactly was exposed.

Similar incidents have happened before in the software industry, from leaks of operating system source code to leaks of code from individual services. Experience shows that the real damage depends not only on the volume of data exposed but also on how fast the company responds: how quickly it can limit further spread of the material, close potential vulnerabilities and restore the trust of customers and partners.

What's next for Anthropic?

Bloomberg's wording that the company is "rushing" to contain the leak suggests the incident had not been fully resolved at the time of publication. Anthropic is likely simultaneously removing or restricting access to the exposed material, assessing which part of the code became public, and analyzing the possible security consequences for the product. For a company whose business model largely rests on enterprise customers' trust in the closed nature and security of its agentic tools, the speed and transparency of its response to such an incident will matter no less than the fact of containing the leak.

It is also telling that the incident affects the very Anthropic product that has recently been presented as the company's main driver of commercial growth, not a side experiment: the more revenue a product generates, the higher the price of reputational or technical costs from any incident involving it. For now, Bloomberg Tech does not disclose details of how the leak happened or how much code was affected, but the very wording "rushing to contain" indicates that the company regards the situation as one requiring an immediate response rather than a scheduled one.