Skip to playerSkip to main content
  • 9 hours ago
Anthropic contained a leak of internal Claude Code instructions after a packaging error exposed source code on GitHub, revealing proprietary AI coding agent techniques. The company issued 8,000+ copyright takedowns but copies spread rapidly. No customer data or model weights were exposed. The incident comes as Anthropic is valued at $380B ahead of a potential IPO.
Transcript
00:00It's Benzinga, bringing Wall Street to Main Street.
00:02Anthropic moved to contain a leak of internal cloud code instructions
00:06after a release packaging error exposed source code on GitHub,
00:09according to the Wall Street Journal.
00:11The company said the leak did not include customer data or model weights.
00:15The exposed materials revealed proprietary techniques,
00:18tools, and instructions used to guide its AI models as coding agents.
00:23Developers and competitors gained access to detailed methods
00:26that could be replicated without reverse engineering.
00:29Anthropic issued copyright takedown requests that removed more than 8,000 copies,
00:34but new versions spread quickly.
00:36The leak also created risks that hackers could probe the code for vulnerabilities.
00:40The incident comes as it secures funding,
00:43valuing the company at $380 billion ahead of a potential IPO.
00:47For all things money, visit Benzinga.com.
Comments