Goldman took Claude Code off its bankers — weeks before Anthropic's IPO

Goldman Sachs restricted Anthropic's Claude Code and Microsoft's GitHub Copilot to its roughly 12,000 engineers, pulling access from the bankers and traders who had used both tools for months, Bloomberg reported on November 2. Traders had been using the coding agents to monitor client trading flows and to draft written and graphic material — jobs with no software-development need.

The reversal lands three months after Goldman's chief information officer Marco Argenti publicly praised the same workflow — a trader building a quick AI prototype and handing it to an engineer to scale — and weeks before Anthropic's IPO, a listing Goldman is lead-underwriting. A Goldman spokesperson said the bank ran a period of experimentation and is now putting the tools in the hands of those best suited to use them.

The bank is not quitting AI: its own GS AI Assistant, which runs Claude behind the firm's firewall, stays available to everyone. What lost its pass is narrower — third-party agents that write and run code, sitting on trading desks without the review gates engineering teams already have. Some traders were reportedly frustrated to lose it.