Goldman limits AI coding tools to engineers

Goldman Sachs The Goldman Sachs Group, Inc. ↗ 4.67% has stopped most of its staff from using advanced AI coding tools after reviewing productivity and risk concerns, and the decision has annoyed some employees, Bloomberg reported on Monday, November 2. For months the bank let bankers use Anthropic's Claude Code and Microsoft's ↗ 2.75% GitHub at work. Internal guidance updated in October and seen by Bloomberg News now limits those tools to engineers.bloomberg.com

The pullback is the latest sign that Wall Street is still working out rules for how employees use AI. Banks are spending heavily on the technology while also managing cybersecurity, cost and productivity problems, according to a translation of the Bloomberg report.finance.sina.cn

Traders lose a shortcut

Some traders are frustrated by the ban. They had used the tools to speed up tasks such as monitoring client trading flows and modeling market moves, Bloomberg Law reported. Others used them to produce text and charts, according to the same translated report. The new rules do not apply to the bank's "strats," staff whose work combines trading, quantitative research and software engineering, the report said.news.bloomberglaw.com+2

Goldman spokesperson Nick Carcaterra said the bank opened some AI tools to employees so it could find the best uses across its businesses. After a trial period, he said, it weighed efficiency, risk control and other factors and gave access to the people best placed to use the tools over the long term, according to the translated account.finance.sina.cn

A shift from vibe coding

The restriction is a change of tone from July. That month, Chief Information Officer Marco Argenti praised the idea of non-technical staff such as traders and bankers "vibe coding," building rough prototypes with AI before handing them to the firm's roughly 12,000 engineers to finish, the report noted.finance.sina.cn

This is also not the first time Goldman has limited AI access. Earlier this year it barred Hong Kong staff from using Anthropic's Claude, according to the report. Major US banks including JPMorgan Chase ↗ 2.27% have also limited employees outside the US from using advanced AI tools like Claude, The Economic Times reported.capitalfutures.com.tw+1

Security worries across the industry

Security has become a common theme for bank technology chiefs. Speaking at the Evident AI Symposium in October, Argenti said getting AI agents to follow company standards and policies remains a key challenge, Banking Dive reported. At the same event, JPMorgan CIO Lori Beer described the tension many banks now face. "We want to move at speed, but sometimes we have to pull back to get controls in place," she said. "The technology is running ahead of capabilities to manage it."bankingdive.com