What I learnt co-leading an AI Safety bootcamp for legal and governance practit

AI safety has become a hot topic, with increasing interest in research and policy developments across the US, UK, and Europe. However, the bottleneck lies in political will and market incentives, making it challenging for policy officers to ensure proper implementation. Companies driven by profit can be difficult to steer, but their internal personnel can still make a significant impact by introducing tighter contractual clauses, enhancing vendor scrutiny, and reducing non-compliance risks.

The EU, despite lacking native frontier labs, remains a substantial market with a regulatory body dedicated to AI safety.

To address this, ML4Good and EquiStamp organized a five-day residential bootcamp in East Sussex for 19 senior legal, governance, and risk professionals from 12 different jurisdictions. The bootcamp aimed to provide participants with a comprehensive understanding of the frontier AI pipeline, from training and evaluations to agents and safeguards.

The unique aspect of this event was the diverse audience, comprising experienced professionals from various backgrounds, including lawyers, privacy and compliance experts, risk officers, financial professionals, legal scholars, and research fellows, as well as representatives from the judiciary and public sector.

The co-leader of the bootcamp, along with Elsa Donnat, emphasized the importance of practical exercises during the course, with technical sessions led by Douw Marx and Jeff Iuliano. These experts had to simplify complex technical concepts for a non-technical, mostly legal audience. The personal and professional realizations derived from this experience included the isolation senior professionals may face when addressing AI safety concerns and the value of building a community among these individuals.

The community aspect of the bootcamp proved to be one of its most significant assets, fostering friendships, connections, and ongoing conversations among participants. One of the most influential attendees expressed gratitude for the event, stating, "Thank you for this. I have found my people." This sentiment was echoed by others who, despite their expertise, were seeking a broader understanding of AI safety issues.

During the bootcamp, participants engaged with various AI safety research, such as the Hugging Face incident, the Redwood Research AI Control paper, and the Golden Gate Claude work, deepening their appreciation for the importance of these problems.