Anthropic is learning the hardest lesson in tech: once code hits the internet, it belongs to the internet.

The Wall Street Journal reports that Anthropic has issued more than 8,000 copyright takedown requests to remove copies and adaptations of the Claude Code source code that leaked last week through npm. The company accidentally shipped its entire 512,000-line codebase to the public registry, and developers did what developers do: they forked it, mirrored it, and started picking it apart before Anthropic even noticed.

Here is the problem: 8,000 takedowns is not a containment strategy. It is whack-a-mole with a legal mallet.

Systima, an AI agent consultancy, put it bluntly: "The source is, for all practical purposes, permanently public." Developers are already rewriting and restructuring the leaked code specifically to dodge copyright claims. The underlying logic, architecture decisions, and system prompts that power Claude Code are now common knowledge in the open-source community.

The silver lining, if you squint: the leak did not expose model weights, training data, or API infrastructure. This was the instruction layer, not the brain. But for a company that has been positioning itself as the responsible, buttoned-up alternative to OpenAI, accidentally publishing your entire product to npm is not exactly a trust-building exercise.

The Copyright Irony Nobody Is Talking About

Here is where it gets interesting. Anthropic has bragged publicly that Claude writes a significant portion of its own code. Reddit and Slashdot users are already asking the obvious question: if an AI wrote the code, does Anthropic actually hold the copyright? US copyright law requires human authorship. If Claude Code was substantially authored by Claude itself, Anthropic may be filing takedown requests on code it cannot legally own.

That is not just an embarrassing legal footnote. It is a potential precedent that could reshape how every AI company protects its intellectual property. The entire industry is racing to have AI write its own tools. Nobody has figured out who owns the output.

What Happens Next

Anthropic will keep filing takedowns. Developers will keep forking. The code is out there, permanently, and the only real question is whether competitors will use it to build better products. Meanwhile, Anthropic is reportedly blaming human employees for the leak, which is the oldest play in the corporate crisis handbook: when the system fails, find a person to blame.

For a company valued at $61.5 billion and growing, this is survivable. But the combination of the leak, the copyright questions, and the blame game is not a great look for the company that wants you to trust it with the future of AI safety. First reported by the Wall Street Journal.