The biggest obstacle to safer AI is embarrassment

Embarrassment stands as the principal barrier to safer AI, as firms are often reluctant to disclose AI-related mistakes. This reluctance arises from the fear of negative consequences that could arise from admitting such errors. To overcome this hurdle, the financial services industry could emulate the Aviation Safety Reporting System, a model that has been in place for decades.

This system encourages the voluntary and confidential reporting of near misses, which are then analyzed and shared among the industry. By creating a confidential repository for reporting AI incidents and near misses, financial services can learn from each other's experiences and improve overall safety. However, the core challenge lies in overcoming cultural resistance to sharing this information, which is deeply rooted in the fear of embarrassment and the desire to protect competitive advantage.

Eliminating embarrassment through measures like Chatham House Rules, where the origin of the information cannot be traced, could help unlock the full potential of such a repository. Ultimately, the key to a safer AI landscape is fostering a culture where firms feel comfortable sharing their AI experiences, turning potential embarrassment into opportunities for collective improvement.