AI Company Leaks Its Own Genius Coding Formula—Because Humans

KEY POINTS

•Anthropic accidentally leaked internal source code for its Claude Code AI assistant in a packaging error confirmed in March 2026.
•The leak did not involve any sensitive customer data or credentials, said an Anthropic spokesperson calling it a 'human error.'
•Following a public rift with the Pentagon in February, Anthropic's Claude chatbot surged to number one in the U.S. App Store.

Anthropic, famed for its AI coding assistant Claude Code, somehow managed to release internal source code by accident in early March 2026—not dark web style, but corporate oops worthy of a sitcom. CEO Dario Amodei, fresh off a February breakup with the Pentagon over AI usage, watched as his company rose to No. 1 in the U.S. Apple App Store even while leaking code to rivals and awarding X a cool 26 million views on the spill. Anthropic insists no customer secrets were exposed, chalking it up to 'human error'—because nothing says Silicon Valley like handing your proprietary coding recipe out like Halloween candy.

Share the Story

(1 of 3)

Mockingbird News

@MockingbirdNews · now

Anthropic gave its competitors free tutoring by accidentally tweeting out the source code. AI safety feels quaint when your biggest threat is your own release manager.

184/280 charactersReady to post

Mockingbird News

@MockingbirdNews · now

When your AI safety policy is 'try not to leak code,' you might want to rethink hiring the intern responsible for February's Pentagon breakup and March’s code spill.

182/280 charactersReady to post

Mockingbird News

@MockingbirdNews · now

Anthropic accidentally shared Claude Code’s secret sauce publicly, proving even AI companies can’t keep their own secrets better than your uncle’s embarrassing texts.

183/280 charactersReady to post