AI TRENDS | Goldman Sachs Restricts Traders and Investment Bankers’ Use of Some AI Tools

Goldman Sachs has barred most employees from using advanced AI coding tools, restricting access to engineers after assessing efficiency and risk. According to Sina Finance, citing Bloomberg, an October update to internal guidance narrowed access to Anthropic’s Claude Code and Microsoft’s GitHub, which investment bankers had been allowed to use in recent months. An unnamed source said some traders were frustrated by the ban. They had used the tools to monitor client transaction flows, build models tracking markets and position changes, and create written materials and charts. In July, Goldman Sachs Chief Information Officer Marco Argenti praised non-technical employees, including traders and investment bankers, for using “vibe coding”: they could build initial solutions with AI tools and then have the company’s 12,000 engineers refine them. Goldman spokesperson Nick Carcaterra said the firm gives employees access to selected AI tools to explore their best uses across business lines. After a trial period and an assessment of efficiency, risk controls and other factors, the firm assigned access to employees best suited to use the tools over the long term. The new rules do not apply to strats, roles spanning trading, quantitative research and software engineering, another source said. The change is another shift in Wall Street’s efforts to set rules for employee AI use, as banks invest in AI while addressing cybersecurity, cost and productivity challenges. In July, JPMorgan said it would guide employees to choose models according to their needs and avoid using expensive large models for simple tasks such as summarizing research reports. Goldman has also developed its own AI assistant, integrating models from OpenAI, Anthropic and Google. Goldman is lead underwriter for Anthropic’s IPO and has partnered with chipmaker Nvidia on several large bond financings. In May, John Waldron, viewed as a potential successor to CEO David Solomon, described Goldman as a “human assembly line” with significant potential for automation. Goldman had also previously restricted Hong Kong-based employees’ use of Anthropic’s Claude.