Skip to content
Industry · Nov 8, 2026

Goldman Sachs restricts third-party AI coding agents for most staff, citing risk and productivity concerns

Bank shifts access to engineers weeks before leading Anthropic’s IPO, as internal review flags unreviewed code and information barrier risks.

Trust75
HypeLow hype

2 sources · cross-referenced

ShareXLinkedInEmail
TL;DR
  • Goldman Sachs has restricted third-party AI coding agents like Anthropic’s Claude Code and Microsoft’s GitHub Copilot to engineers only, removing access for most traders and bankers.
  • The change follows an internal review citing productivity and risk concerns, including unreviewed code and information barrier compliance.
  • The restriction applies even as Goldman serves as a lead underwriter for Anthropic’s upcoming IPO, expected before the end of November.
  • Goldman previously removed Claude Code from its Hong Kong platform in April, citing contract terms.

Goldman Sachs has restricted access to third-party AI coding agents—including Anthropic’s Claude Code and Microsoft’s GitHub Copilot—to its engineering staff, effectively cutting off most traders and bankers from using the tools at work. The bank confirmed the change in an October internal guidance update following a review of productivity and risk concerns, Bloomberg reported.

The restriction applies to autonomous coding agents that can write and execute code independently, though Goldman’s in-house AI assistant, which uses models from multiple providers, remains available across the firm. Goldman’s strats—employees who work across trading, quantitative research, and software engineering—retain access to the third-party agents, according to a person familiar with the matter.

The decision has frustrated some traders who had used the tools to monitor client trading flows, model market positions, and generate written and graphic materials, a person familiar with the matter told Bloomberg. Previously, non-engineering staff had built tools with AI coding agents and handed them to engineers—about 12,000 of Goldman’s roughly 47,000 employees—for refinement and deployment.

Goldman’s chief information officer, Marco Argenti, had publicly praised this ‘vibe coding’ approach as recently as July, noting that traders and bankers could prototype tools and pass them to engineers for improvement. The October guidance limits such agent use to engineers only, aligning with the bank’s broader AI adoption strategy described in waves, with developers as the first wave.

The bank had already restricted Claude Code in Hong Kong in April, citing contract terms with Anthropic; an Anthropic spokesperson said Claude had never been officially supported in Hong Kong, while ChatGPT and Gemini remained available there. JPMorgan removed Claude from its Hong Kong approved list in June. Goldman is now applying role-based restrictions firmwide, weeks before it serves as a lead underwriter on Anthropic’s IPO, which is expected to list on Nasdaq under the ticker ANT before the end of November.

Goldman’s internal review flagged risks including unreviewed code reaching production and potential breaches of information barriers—controls that separate teams to prevent insider trading. Engineers typically review, test, and deploy code before it reaches critical systems, but agent-built scripts can enter daily use without such oversight. Goldman’s internal AI platform was designed with these controls in mind, but tools built by non-engineers using autonomous agents were not subject to the same safeguards.

Cost may have also factored into the decision. Usage of agent tools is billed by the token, and several companies have seen spending rise as employees run agents on tasks that do not align with business priorities. Goldman’s move follows similar restrictions at other financial firms this fall, including JPMorgan and Schwab, which have blocked outside agents from customer accounts.

Sources
  1. 01State of AI — Goldman Sachs Is Pulling Claude Code from Its Bankers and Traders
  2. 02Bloomberg News — Goldman Sachs Restricts AI Coding Tools for Most Staff, Citing Risk and Productivity
Also on Industry

Stories may contain errors. Dispatch is assembled with AI assistance and curated by human editors; despite the trust-score filter, mistakes happen. We correct publicly — every article links to its revision history. Nothing here is financial, legal, or medical advice. Verify before relying on any claim.

© 2026 Dispatch. No ads. No sponsorships. No paid placement. Reader-supported via Ko-fi.

Built by a person who cares about honest AI news.