Overview

• As of July 23, the generative AI assistant Elsa remains active in FDA review processes but continues to fabricate studies and misrepresent data
• Internal users say Elsa confidently gives incorrect drug-approval answers and invents nonexistent research, forcing extra manual fact-checking
• Commissioner Marty Makary deployed Elsa agency-wide on June 2 under budget and hailed its cost-effectiveness while acknowledging potential hallucination risks
• HHS Secretary Robert F. Kennedy Jr.’s push for AI in health policy was highlighted by a MAHA commission report that cited at least seven studies that did not exist
• Agency leaders are under pressure to balance AI-driven acceleration of drug approvals with enhanced oversight measures and policy guardrails to safeguard integrity