Goldman Sachs has restricted the use of several advanced AI coding tools among its employees, limiting access primarily to engineers.
The decision affects applications such as Anthropic’s Claude Code and Microsoft’s GitHub, which some traders and bankers had previously used to speed up everyday tasks.
According to Bloomberg, the bank introduced the restrictions in internal guidance updated in October after reviewing the tools’ impact on productivity and the risks associated with their use. The change has frustrated some traders who relied on AI to monitor client trading activity, model market movements and positions, and prepare written and visual materials.
The restrictions mark a shift from the approach Goldman had been taking earlier this year. In July, Chief Information Officer Marco Argenti spoke positively about the use of so-called “vibe coding” by employees without technical backgrounds.
Traders and bankers were experimenting with AI tools to build their own applications and then passing the results to the bank’s engineering teams for further development. Goldman employs around 12,000 engineers.
A company spokesperson said the bank had initially made certain AI tools available to employees so they could identify useful applications across different business areas. After testing them for productivity and risk management, Goldman decided to restrict long-term access to employees best positioned to use the technology.
The move reflects a broader challenge facing Wall Street banks. Financial institutions are investing heavily in AI while trying to control spending, address cybersecurity concerns and determine where the technology actually improves productivity.
JPMorgan Chase, for example, said in July that it wanted employees to choose AI models according to the task at hand. This approach is intended to prevent staff from using expensive models for relatively simple jobs, such as summarizing research reports. Goldman has also developed its own AI assistant, which combines models from OpenAI, Anthropic, and Alphabet.
The bank’s position on AI is partially complex. Goldman is working with major technology companies and has been involved in financing the sector. It was named a lead underwriter for Anthropic’s planned initial public offering and has partnered with chipmaker Nvidia on large debt transactions.
In May, John Waldron, who is widely viewed as a potential successor to CEO David Solomon, described Goldman as an “assembly line” with significant potential for automation.
The latest restrictions are not the bank’s first attempt to limit employee access to AI tools. Earlier this year, Goldman also restricted the use of Anthropic’s Claude among staff in Hong Kong. The new guidance does not cover so-called strats, employees whose roles combine trading, quantitative research and software engineering.
