Anthropic is back in the news again, and not for the best reason. The company accidentally exposed part of the source code behind its Claude AI. It happened just days after another small slip involving its upcoming “Claude Mythos” project.
This latest issue picked up quickly online. A post sharing the leaked code started circulating on X and went on to get tens of millions of views. That alone turned what might’ve been a quiet mistake into something much bigger.
According to Anthropic, this wasn’t a hack or outside attack. They say it came down to a release packaging mistake, basically, human error during the process. The company also made it clear that no user data or sensitive credentials were exposed, which is at least some relief.
Then again, the leak itself was no small matter. The archive reportedly consisted of approximately 2,000 files and almost 500,000 lines of code. Soon afterwards, copies also showed up on GitHub.
Even in a case where no personal data was at stake, such incidents habitually draw scrutiny. A little view behind the curtain, okay? Internal code escaping can give folks a closer look at how things operate.
And because something rather like this has happened before, it’s also beginning to feel less like a one-off error and more like something Anthropic needs to get under control.
