Any use of AI or automation in government, particularly as it relates to benefit determinations or claims, needs to be accompanied by policy guardrails that create strong disincentives for improper denials of claims for benefits.