AI company Anthropic is working to contain a leak of its Claude AI agent's source code, with its CCO, Paul Smith, revealing that 'process errors' during rapid product releases caused the incident, not a hack.