Goldman Sachs has barred most traders and bankers from using AI coding tools including Anthropic’s Claude Code and Microsoft’s GitHub, following a review of productivity and risk management. Updated guidance issued in October limits access mainly to engineers, while exempting Strats, employees whose roles combine software engineering, quantitative research and trading. The restrictions have frustrated some traders who used the tools to track client activity, simulate market and position changes, and draft text and visuals. Goldman previously encouraged experimentation, including “vibe coding” by nontechnical staff, but its Chief Information Officer Marco Argenti said the resulting work was passed to engineers from the bank’s 12,000-person pool for refinement. Goldman partner Chris Churchman has warned that widespread AI adoption could weaken the analytical skills of future bankers. The bank is also rolling out its proprietary AI assistant to about 46,000 employees, drawing on models from OpenAI, Anthropic and Alphabet, and has tested agentic AI for functions including trade surveillance. Goldman earlier restricted Anthropic’s Claude for staff in Hong Kong. JPMorgan Chase has separately advised employees to match tasks with appropriately priced AI models.