David Solomon argues AI needs an oversight model beyond banking

Solomon’s central point is that AI should not simply inherit the oversight model used in financial services.[2] He says financial regulatory structures were cobbled together over a century, while AI technology needs scrutiny of its own.[2] The argument does not dismiss risk-management practices from banking: Solomon says Goldman Sachs’ principles could be applied to AI oversight.[1]

Solomon’s proposed approach focuses on how organizations manage the systems they use and the risks they inherit. He says individual decision-makers should be held accountable for the outcomes their systems produce.[1] He also recommends that companies understand risks coming from model developers and routinely stress-test AI systems.[1] Solomon says employees should be rewarded for raising concerns about AI vulnerabilities.[1] Together, the proposals point to oversight as an ongoing organizational responsibility.

Solomon says his proposed principles could apply both to AI developers and to companies that deploy AI.[1] That scope matters for teams deciding where to place oversight: organizations that build models and those that put them to work may need to consider their own risks and responsibilities. The article’s emphasis on risks inherited from model developers also makes the relationship between those groups part of the oversight question.[1] Solomon’s stated principles offer a starting point for discussion, not evidence that any particular implementation or control has been adopted.

Solomon’s argument comes alongside concerns about the AI buildout that he acknowledged during Goldman Sachs’ third-quarter earnings call, including financial concerns and safety worries.[4] He also called for stronger guardrails for AI.[4] These points connect the oversight discussion to potential consequences beyond financial services, while the article’s core recommendation remains that AI requires scrutiny suited to its technology.