Artificial intelligence in FDA regulatory correspondence, 2022 to 2026: a retrospective document analysis

This review measured how often, and in what context, the U.S. Food and Drug Administration (FDA) refers to artificial intelligence (AI), including generative AI, in warning letters. A search was conducted of the full text of all FDA warning letters issued from January 1, 2022, through September 28, 2026, taken from FDA’s public warning letter database. The search terms were “artificial intelligence,” “AI,” “machine learning,” and 4 related terms. Every letter containing a term was assessed and read in full, and the violations cited in each included letter were counted. Form FDA 483s and establishment inspection reports (EIRs) that refer to AI were described separately. Of 3,011 warning letters, 7 (0.23%) referred to AI: 3 device, 3 tobacco, and 1 drug letter. Only the April 2026 letter to Purolea Cosmetics Lab treated a firm’s use of AI as a violation, citing unreviewed AI-generated documents under 21 CFR 211.22(c). In addition, 1 EIR and 2 Form 483s referred to AI. FDA references to AI in warning letters are rare, recent, and heterogeneous, and most describe AI-enabled products subject to existing authorization requirements. The Purolea letter shows that FDA will cite unreviewed AI output under existing current good manufacturing practice (CGMP) rules.