Goldman limits Claude Code to engineers, cutting off traders

Goldman Sachs has limited Anthropic’s Claude Code and Microsoft’s GitHub to its engineers, which cuts off most of its bankers and traders. The change comes from internal guidance updated in October and seen by Bloomberg News, which reported it on November 2. The rule is based on job role and covers third-party agents that can write and run code by themselves. Staff outside engineering lose those agents but keep Goldman’s in-house AI assistant, which runs on models from several companies and is still available across the firm.

Strats, who work across trading, quantitative research and software engineering, also kept access, a second person familiar with the matter said. Developers make up more than 12,000 of the bank’s roughly 47,000 staff, so about a quarter of the workforce keeps the agents, plus the strats. The guidance followed a review of productivity and risk concerns.

What traders used it for

The change has frustrated some traders, a person familiar with the matter told Bloomberg. They had used the tools to monitor client trading flows, model market moves and positions, and put together written and graphic material. Goldman spokesperson Nick Carcaterra said the bank makes certain AI tools available to staff “to determine the best use cases” across the firm.

As recently as July, chief information officer Marco Argenti was talking up “vibe coding” by Goldman’s non-technical employees. He told Bloomberg that traders and bankers had built tools with AI and handed them to the firm’s engineers to improve. Under that arrangement, the people closest to a problem could build a first version without waiting for engineering time. It also meant thousands of employees without engineering training were running software that writes and executes code, on the same machines they use to handle client information.

Argenti’s description of how Goldman’s developers work with coding agents has humans at both ends. A developer sets the specs, regulatory parameters and intended outcome, the agent does the work, humans review the code, and the agent tests and validates it, per American Banker. At a conference in Turin in October, Argenti described Goldman’s AI adoption as a series of waves, with developers as the first. The new guidance keeps the coding agents with that first group.

Why Hong Kong was different

Goldman had already restricted Claude once this year. In April it removed Claude from its internal AI platform for bankers in Hong Kong. The Financial Times reported that the bank had read its contract with Anthropic strictly, and an Anthropic spokesperson said Claude had never been officially supported in Hong Kong. ChatGPT and Gemini stayed available there, and JPMorgan removed Claude from its Hong Kong approved list in June. That restriction was based on location and contract terms. The October one is based on job role and applies across the firm.

Anthropic in the back office

Goldman’s own work with Claude goes on. Anthropic engineers spent six months at the bank helping build Claude-based agents for trade accounting and client onboarding, Argenti said in February, per CNBC. He said the bank was “surprised” at how well Claude handled work beyond coding, especially accounting and compliance. Those agents sit in the back office and are run by Goldman’s engineering side.

Goldman’s other tie to Anthropic is financial. Anthropic’s S-1, filed October 26, names Goldman a lead underwriter on the IPO alongside Morgan Stanley.