David Solomon, chair and CEO of Goldman Sachs, has called for a distinct approach to regulating artificial intelligence (AI), arguing that the technology’s unique characteristics require bespoke oversight rather than borrowing frameworks from other highly regulated industries such as banking.

Solomon highlighted the transformative potential of AI across multiple sectors, citing a recent breakthrough in healthcare where an AI model developed by the Mayo Clinic identified 73 percent of pancreatic cancers up to three years before clinical diagnosis. This early detection capability could significantly improve survival rates for a disease typically diagnosed too late for effective intervention. Beyond medicine, AI promises advances in drug discovery, food production, supply chains, transportation, and national security.

Despite the widespread benefits, Solomon cautioned against rushing unchecked innovation or imposing overly restrictive regulations that could stifle progress. He described the current public debate on AI regulation as trapped in a false dichotomy—either accelerating AI development without guardrails or halting innovation through heavy-handed controls. Instead, he advocated a layered governance approach involving governments setting regulatory frameworks, developers building safe and reliable AI models, and companies deploying these models responsibly within their operations.

Solomon rejected proposals to directly apply banking’s regulatory regime to AI, emphasizing that AI’s rapid evolution, technical complexity, and diverse use cases demand tailored oversight rather than replication. However, he suggested that some risk management principles long used in the financial industry could inform how companies integrate AI, especially as many organizations now see themselves as AI-driven enterprises.

Among these principles, Solomon stressed the importance of accountability, with individual decision-makers held responsible for AI-related risks and outcomes, rather than blaming the technology itself. He also highlighted the need for understanding counterparty risk, urging firms to thoroughly evaluate and monitor AI vendors to avoid overreliance on single providers. This includes ensuring that AI models undergo independent review, thorough documentation, continuous monitoring, and ongoing reassessment as capabilities evolve.

Another key recommendation is routine stress testing of AI systems, using realistic and severe scenarios to reveal potential failures, misalignments, and unintended consequences before widespread deployment. Solomon noted that while banks routinely perform stress tests for regulatory compliance and internal risk management, similar practices are less common in AI but critical for building trust.

Finally, the Goldman Sachs CEO advocated fostering a corporate culture that encourages employees at all levels to escalate concerns about AI vulnerabilities. Recognizing and rewarding risk escalation, he argued, is essential to maintaining safe and trustworthy AI deployment.

Solomon concluded that the companies that succeed in the AI era will not be those that move the fastest, but those that develop disciplined, accountable, and confidence-inspiring approaches to scaling AI technology. The challenge, he said, is not to choose between innovation and safety but to avoid poorly governed experimentation and instead pursue progress grounded in responsibility and trust.