Stephen Casper - ML Researchers as Policymakers [Alignment Workshop]

Stephen Casper (MIT CSAIL) demonstrates how ML researchers can directly influence policy by strategically conducting technical research that operationalizes deliberately vague legislative terms. Lawmakers globally are punting technical decisions to the ML community through terms like 'reasonable safeguards,' 'foreseeable risks,' and 'state-of-the-art techniques.' Casper's deepfake research exemplifies this approach: by documenting how a small number of model developers represent critical bottlenecks in harm supply chains, and showing that basic mitigations create order-of-magnitude differences in abuse patterns. The work directly informed Arkansas HB 1529—legislation allowing the Attorney General to sue developers lacking 'reasonable safeguards.' This creates a reproducible pathway: identify vague legislative terms, conduct empirical work operationalizing them, then communicate findings directly to policymakers and legal stakeholders. Note: The opinions shared in this event are those of the speaker(s) and may not represent the views of FAR.AI or their affiliated organizations.