FDA warning letter links AI use to drug‑manufacturing quality lapses, legal analysis warns
- The Food and Drug Administration’s April 2 warning letter to Purolea Cosmetics Lab said the Michigan drug maker used artificial intelligence to draft required manufacturing records without adequate human review. - FDA said Purolea used AI agents to create specifications, procedures and master production records, then shipped drug products without required process validation after staff said the tool never flagged the rule. - The letter applies existing current Good Manufacturing Practice rules to AI use, extending FDA’s January 2025 draft AI guidance from submissions into factory-floor quality controls. (dlapiper.com)
Drugmakers just got a blunt FDA message: artificial intelligence can help write records, but it cannot take over quality decisions. (fda.gov) (jdsupra.com) The warning letter was issued April 2, 2026, to Purolea Cosmetics Lab after an FDA inspection, and DLA Piper flagged it on April 24 as the agency’s first drug-manufacturing warning letter explicitly tied to AI misuse. (dlapiper.com) (gmp-compliance.org) Current Good Manufacturing Practice rules are the factory playbook for drugs: they require written procedures, validated processes, and a quality unit that signs off on records affecting product quality. (fda.gov) (jdsupra.com) FDA said Purolea used AI agents to generate drug product specifications, procedures, and master production records, which are the step-by-step instructions used to make each batch. (dlapiper.com) (raps.org) When investigators asked about missing legal requirements, personnel said they did not know some obligations existed because the AI agent they relied on did not tell them. (dlapiper.com) (redica.com) One missing step was process validation, the evidence that a manufacturing process can reliably make the same drug quality every time before products are distributed. FDA said Purolea distributed drug products without first performing that validation. (jdsupra.com) (redica.com) FDA tied the problems to 21 CFR 211.22(c), which makes the quality unit responsible for approving procedures and specifications, and 21 CFR 211.100, which requires written production and process controls. (jdsupra.com) (gmp-compliance.org) The agency did not say AI is banned in drug manufacturing. The point was narrower: AI-generated documents and recommendations still have to be reviewed and cleared by an authorized human in the quality unit. (dlapiper.com) (redica.com) That fits with FDA’s January 6, 2025 draft guidance on using AI to support regulatory decision-making for drugs and biologics, which described a risk-based framework for judging whether an AI model is credible for quality, safety, or effectiveness decisions. (jdsupra.com) The warning also lands after FDA’s own December 1, 2025 announcement that it had deployed agentic AI tools for agency employees, underscoring that the regulator is adopting AI internally while policing how industry uses it. (fda.gov) For manufacturers, the lesson in the letter is old-fashioned even if the software is new: if an AI system writes the record, a human still owns the signature, the validation, and the batch. (dlapiper.com)
