RC RANDOM CHAOS

Goldman Sachs Restricts Claude Code to Engineers Only

· via Hacker News

Original source

Goldman Sachs is pulling Claude Code from its bankers and traders

Hacker News →

Goldman Sachs restricted Anthropic’s Claude Code and Microsoft’s GitHub to engineers only in October, cutting off traders and bankers who used the tools to track client flows and build market models. The bank’s roughly 12,000 engineers retain access, as do Strats working across trading, quantitative research, and software engineering. The move reversed a July stance by CIO Marco Argenti, who had highlighted non-technical employees building initial tools with AI for engineering teams to refine.

The restriction addresses operational and regulatory concerns. When traders use Claude Code, generated tools run with user-level access but lack the review, testing, and deployment controls applied to engineer-written software. Traders modeling positions with the tool may not fully understand the generated code, creating risk on a trading floor. Tools built outside engineering bypass information barriers that regulators require between different business units. Token-based billing also contributed; agent usage costs have climbed at other firms, with Accenture allocating $29 monthly per employee. Goldman’s internal AI assistant using multiple models remains broadly available, and back-office agents for trade accounting and client onboarding built with Anthropic engineers continue operating under engineering teams.

Goldman and Morgan Stanley lead Anthropic’s expected Nasdaq IPO under ticker ANT before the end of November. Barclays is deploying Claude Code to half its developers while cutting 5,000 staff. JPMorgan removed Claude from its Hong Kong approved list in June, and JPMorgan and Schwab block outside agents from customer accounts.

Read the full article

Continue reading at Hacker News →

This is an AI-generated summary. Read the original for the full story.