Goldman Sachs has restricted access to advanced artificial intelligence coding tools, including Anthropic’s Claude Code and Microsoft’s GitHub, to its engineering employees. The decision affects thousands of non-technical employees, including traders and bankers who had previously used these tools to automate tasks, analyse financial data and develop software solutions.
The move marks a shift in the investment bank’s approach to AI adoption, following an extended period of experimentation and an internal assessment of productivity and risk management.
Goldman Sachs Limits Access to AI Coding Tools
According to internal guidance updated in October, Goldman Sachs has limited the use of Claude Code and GitHub to engineering teams. The restrictions apply to most non-technical employees who previously had access to these applications.
Traders and bankers had been using the AI-powered coding tools for several activities, including monitoring client trading flows, analysing market movements, modelling financial positions and creating written and graphical content.
The restrictions have reportedly frustrated some employees who had incorporated these tools into their daily workflows.
However, the new policy does not apply to every employee outside traditional engineering teams. Specialist employees known as ‘strats’, who work across trading, quantitative research and software engineering, are exempt from the restrictions.
Earlier Push Towards AI-Powered Development
The decision represents a change from Goldman Sachs’ earlier encouragement of wider AI adoption across the organisation.
In July, Chief Information Officer Marco Argenti had spoken positively about non-technical employees experimenting with AI-assisted software development, commonly known as vibe coding.
Under this approach, employees without formal programming backgrounds could use AI tools to develop basic applications and prototypes. These initial solutions could then be handed over to the bank’s approximately 12,000 engineers for further development and refinement.
The initiative allowed employees across different departments to explore potential applications of AI without requiring extensive programming knowledge.
Productivity and Risk Management Behind the Decision
A Goldman Sachs spokesperson explained that the bank provides selected AI tools to employees to identify suitable applications across different business functions.
The spokesperson said the company evaluates several factors, including productivity and risk management, before deciding which employees should receive long-term access to particular tools.
The decision reflects broader concerns among financial institutions about deploying advanced AI applications in environments where sensitive financial information, cybersecurity and operational accuracy are important.
Banks are increasingly investing in AI technologies while evaluating their costs, security implications and measurable business benefits.
Wall Street Rethinks Wider AI Adoption
Goldman Sachs is not the only financial institution reassessing how employees use AI tools. JPMorgan Chase has also encouraged employees to select AI models according to the complexity of their tasks, helping avoid unnecessary spending on expensive models for routine activities.
Goldman Sachs has separately developed an internal AI assistant that combines models from OpenAI, Anthropic and Google.
The latest restrictions highlight how financial institutions are moving towards more controlled AI deployment after initial experimentation. Rather than providing unrestricted access, banks are increasingly evaluating which employees and business functions can benefit most from specialised AI tools.
Summary
Goldman Sachs has restricted Claude Code and GitHub access to engineering employees following an assessment of productivity and risk management. The decision affects thousands of traders and bankers who previously used these tools for financial analysis and automation. Specialist employees known as strats remain exempt. The move reflects Wall Street’s growing focus on controlled AI adoption and operational security.
