Goldman Sachs restricts traders and bankers from using Claude Code and GitHub Copilot

What did Goldman Sachs change?

Goldman Sachs issued internal guidance in October that restricts Anthropic's Claude Code and Microsoft's GitHub Copilot to the bank's engineering staff only, according to Bloomberg as reported by Quartz and PYMNTS. The tools had been available to traders and bankers for months while the bank assessed their impact.

A Goldman spokesperson told Bloomberg that the bank makes certain AI tools available to its people to determine the best use cases, and after a period of experimentation assessing productivity and risk management, the tools are put in the hands of those best suited to leverage them for the long term.

The internal guidance was published in October and seen by Bloomberg. The move represents a notable reversal for Goldman, which had positioned itself as one of Wall Street's more aggressive adopters of generative AI tools across its workforce.

How did traders react?

The restriction has frustrated some Goldman traders, according to Bloomberg. They had adopted the AI coding tools for tasks such as following how clients were trading, simulating shifts in markets and in their positions, and drafting text and visuals.

The directive does not cover "strats," a group of employees whose jobs combine software engineering, quantitative research and trading, according to Bloomberg. The carve-out suggests the bank is distinguishing between AI-assisted coding for engineering work and AI-assisted financial analysis.

The frustration among traders is understandable: they had integrated Claude Code and GitHub Copilot into their daily workflows, using the tools to automate repetitive coding tasks and accelerate data analysis. The restriction forces them back to manual processes or to queue requests through engineering, adding latency to time-sensitive trading operations.

Why did Goldman reverse its AI policy?

Goldman's reversal from CIO Marco Argenti's July endorsement of "vibe coding" to the October restriction reflects a pattern seen across Wall Street. In July, Argenti said non-technical employees used AI to create their own solutions, then handed results to one of the bank's 12,000 engineers for polishing.

But Goldman partner Chris Churchman warned in August that over-reliance on AI could erode the analytical skills of junior bankers, who traditionally learn by doing the detailed work that AI now automates. Churchman, who co-chairs the bank's Global Banking and Markets AI working group, said Goldman had not yet determined how it would manage that transition, according to Bloomberg.

The Hong Kong restriction on Claude earlier this year suggests regulatory compliance was also a factor. Financial data in Hong Kong is subject to cross-border data transfer rules that complicate the use of cloud-based AI tools.

The reversal highlights the tension between encouraging AI experimentation and maintaining control over proprietary financial data and compliance obligations.

How does this compare to other banks?

Goldman is not alone in drawing boundaries around AI. JPMorgan Chase CEO Jamie Dimon warned of "rogue agents" during the bank's third-quarter earnings call and said outside AI agents would not be permitted inside the bank, according to PYMNTS.

Bank of America took a conditional approach, limiting AI agents' access to consumer deposits while developing protocols for liability on AI-authorized payments. Citigroup drew the same line, barring third-party agents from logging into accounts or transacting, according to PYMNTS.

Goldman has created its own AI assistant, which draws on models from OpenAI, Anthropic and Google, and that assistant remains available to staff across the firm. The bank continues to invest heavily in AI despite the pullback on specific external tools for non-engineers.

What was Goldman's previous AI stance?

Goldman had been relatively permissive with AI tools compared to some peers. CIO Marco Argenti spoke positively about non-engineers using AI to build their own solutions in July, describing a workflow where traders and bankers created prototypes that engineers then polished.

Goldman's AI investments have been significant. The bank has been a major enterprise customer for both Anthropic and OpenAI, and has experimented with AI across trading, risk management, and client services. The bank also launched its own internal AI assistant earlier this year that draws on models from multiple providers.

What does this mean for Wall Street's AI adoption?

Goldman's restriction comes at a pivotal moment for AI adoption in financial services. Banks are spending millions on AI tools while simultaneously wrestling with governance, compliance, and the risk of junior talent losing essential analytical skills.

JPMorgan Chase CEO Jamie Dimon warned of "rogue agents" and said outside AI agents would not be permitted inside the bank. Bank of America placed limits on AI agents' access to consumer deposits. Citigroup barred third-party agents from logging into accounts or transacting.

The pattern suggests a two-tier AI strategy taking shape on Wall Street: AI tools for engineers are expanding, while access for traders and bankers is being scrutinized and in some cases curtailed.