A.I. Rewrites Leaked Code, Testing Copyright Limits
A University of British Columbia student used artificial intelligence to rewrite leaked Anthropic source code, then posted the modified version online where it quickly accumulated over 100,000 shares. The move raises a direct question for creatives: what does copyright protect when A.I. can reproduce any work in minutes?
Sigrid Jin, 25, saw news that Anthropic had accidentally leaked Claude Code, its A.I. agent software. He posted a copy online. His girlfriend's immediate concern-was this copyright infringement?-prompted him to try a workaround.
Jin directed A.I. assistants to rewrite the code in a different programming language, then shared that version. The rewritten code stayed online even after Anthropic demanded thousands of similar posts be removed from GitHub, the code repository site.
Anthropic cited copyright violations in its takedown requests. The company did not respond to questions about why Jin's rewritten version remained untouched or whether the rewrite qualified as a new work outside their copyright claims.
Jin said he believed rewriting the code transformed it into something Anthropic could not claim ownership of. He framed the action as a philosophical statement rather than a grab for attention or money.
"Any creative work can be reproduced in a second," Jin said. He wanted to raise questions about the value of copyrighted intellectual property when A.I. can duplicate art, music, and literature within minutes.
The incident exposes a gap between copyright law and A.I. capability. Courts have not yet ruled on whether A.I.-assisted rewrites constitute new works or derivative copies. For creatives relying on copyright to protect their work, the answer matters.
Learn more about how generative A.I. and large language models work, or explore resources designed for creatives navigating A.I. tools.
Your membership also unlocks:
