Representative Image
Paul Smith, Anthropic’s chief commercial officer, has disclosed the precise cause of the accidental leak of Claude Code’s source code. This comes after Anthropic confirmed that “human error” caused commercially sensitive information about the company’s AI Agent, which wiped trillions from stock markets, to go viral on GitHub. In an interview with Bloomberg, Smith said that the leak was caused by “process errors” tied to the company’s rapid product release cycle and was not a security breach or hack.Addressing the incident, Smith said the accidental exposure of the source code reflected the challenges of moving quickly in a competitive AI landscape, adding that the issues have since been addressed. “They’re part of the incredibly rapid release cycle that we’ve had around Claude Code,” he said.The comments come as Anthropic continues to compete with OpenAI and Google in building advanced AI systems to power agents and other software.Smith suggested that the pace of model releases will increase. “I think the pace of innovation is going to increase. It is a reasonable assumption that model releases will continue to occur, and if anything, they will occur with greater frequency.”The leak comes at a sensitive moment for the company. Anthropic is currently in a legal dispute with the US government over the Pentagon’s decision to declare it a supply-chain risk following a standoff over AI safety guardrails. The company has warned that the labelling could cost it billions in lost revenue. Still, Smith said some customers respect that Anthropic “demonstrates its principles” in its dealings with the US government.As Anthropic contests the Pentagon’s decision, it is also moving ahead with plans to go public as soon as this year, the Bloomberg report added. Smith declined to comment on the timeline for an initial public offering but said Anthropic is “very comfortable that we’re on the right path from gross margin targets and the previous profitability targets that we’ve stated.”The company has a substantial war chest after finalising a $30 billion funding round in February, which can be used to spend on chips, data centers, and talent needed to compete in the global AI race.”We’re very comfortable with our funding situation after the recent round that we concluded,” he said.Smith said the company’s general-purpose AI agent, Cowork, could reach a broader audience than Claude Code, the product that helped scale the startup’s presence in the AI space. Anthropic has seen stronger early adoption for Cowork “in the first few weeks” than it did during a similar period for Claude Code a year ago, Smith noted in the interview.He added that in a typical large company, only about 2% to 5% of employees work in engineering, while Cowork is designed to appeal to “the rest of us.”Cowork was introduced as a “research preview” earlier this year and gained attention online soon after. Unlike Claude Code, which is built for programming tasks, Cowork does not require users to operate through a terminal or command line. It is designed to handle a wider range of tasks, which may make it relevant to a larger set of users.Anthropic has established itself as a key player in AI and competes with companies such as OpenAI, supported in part by Claude Code, which the company recently said generates more than $2 billion in annualised revenue. With Cowork, Anthropic is looking to extend its tools beyond developers and support a broader set of workplace tasks.At the same time, the company is competing with OpenAI and Alphabet Inc.’s Google to develop more advanced AI systems for agents and software. In the interview, Smith said he expects the pace of new releases to increase.
