Goldman Sachs restricted Claude Code for bankers — but made one surprising exception

<finance sage Amjad Masad warns that AI coding agents running on employee devices pose significant security risks for enterprises and governments. Financial institutions, including Goldman Sachs, are responding by restricting access to tools like Claude Code and GitHub coding tools. Goldman Sachs reportedly restricts access to these tools for bankers and traders, but allows certain quantitative employees known as strats to continue using them.

Replit CEO Masad suggests that this is a widespread issue in finance, with "the ban hammer coming down." He attributes the crackdown to concerns over supply chain attacks, credential exposure, malicious instructions, and compromised dependencies. Masad recommends keeping AI agents in company-controlled cloud environments rather than on employee devices, citing concerns about credential exposure, malicious instructions, and compromised dependencies.

Replit provides a security advisory recommending this approach, pointing to recent security incidents as examples of how attackers can exploit developer tools. While restricting AI coding agents to engineers may not fully solve the problem, Masad argues that companies need to control what agents can do, not just who gets to use them.

Goldman Sachs' restrictions indicate a broader trend of companies rethinking how widely they allow employees to use AI coding tools.