Goldman Sachs CEO David Solomon has outlined a set of principles for AI governance, emphasizing that individual decision-makers must be held accountable for the outcomes of the systems they deploy. The framework suggests that companies should routinely stress-test AI, evaluate risks inherited from model developers, and incentivize employees to report identified vulnerabilities.
These guidelines are intended for both AI developers and organizations implementing the technology. Solomon argues that progress should be built on trust and accountability rather than poorly governed experimentation. The bank is also focused on preventing knowledge loss, with executives expressing concerns that overreliance on AI could hinder the professional development of junior traders.
The bank has been actively integrating AI into its operations. In February, Goldman Sachs began deploying autonomous agents based on the Anthropic Claude model to automate accounting, compliance, and operational finance tasks. The firm spent six months embedding Anthropic engineers within its internal teams to co-develop these specialized tools.
The broader regulatory landscape remains active as lawmakers consider new frameworks for frontier AI. Meanwhile, major developers like OpenAI and Anthropic are currently negotiating the scope of access for external evaluators within their labs, following earlier agreements to allow independent oversight of their systems.
