Goldman Sachs limits Claude Code and GitHub tools to engineers, bankers and traders lose access

Bloomberg reports Goldman Sachs changed its internal guidance in October: AI coding tools stay with about 12,000 engineers, while bankers and traders lost access.

Goldman Sachs has restricted access to Anthropic's Claude Code and GitHub's coding tools to its engineers. According to Bloomberg, which other outlets relayed, the change comes from internal guidance updated in October 2026. Bankers and traders, who had used the tools for several months, lost access. The story drew more than 700 points on Hacker News.

Reports say about 12,000 engineers out of roughly 47,000 employees keep access, along with so-called strats, who combine software engineering, quantitative research and trading. Bankers and traders had used the tools to monitor client trading flows, simulate market moves and draft text and visuals. Numbers differ slightly between reports, and one outlet speaks of roughly 10,000 affected people.

The bank gave a general reason. A spokesperson said the firm weighs productivity against risk management and that the experimentation phase is over. Goldman did not name a specific security incident or coding failure as the cause. The shift follows a summer in which chief information officer Marco Argenti praised vibe coding by non-technical staff, and it is reportedly causing frustration inside the firm.

Is this a reason for companies to ban AI agents? No. It illustrates the approach that does not work well: hand a tool to everyone first, then pull it back across the board. An agent that can read files, run commands and reach systems needs clearly scoped permissions, an audit trail and human approval for steps with consequences. That the story involves Claude Code is incidental. The same questions apply to any coding agent or assistant with access to company data.