If AI drafted our label, who is responsible for a wrong word in it?
Our reading: the firms that put the labeling out, not the tool. That means the application holder, and any repacker or relabeler that lists the drug and submits its own labeling. US misbranding law looks at the labeling itself: if it's false or misleading in any particular, or lacks adequate directions or warnings, the drug is misbranded, and our reading is that the law makes no exception for text a tool wrote. The FDA's January 2025 draft guidance on AI does not address AI used for drafting a submission, so our reading is that nothing in it shifts that responsibility.
This page is general information, not legal or regulatory advice.
What US law looks at
Section 502 of the federal Food, Drug, and Cosmetic Act says a drug is misbranded if its labeling is false or misleading in any particular, or doesn't carry adequate directions for use and adequate warnings. The test is about the words on the label. Our reading: whether a person or a model drafted them doesn't change the test.
What the FDA's AI guidance does and doesn't say
The FDA's January 2025 draft guidance on AI to support regulatory decision-making says it does not address AI used for operational efficiencies, such as drafting and writing a regulatory submission, when that use doesn't affect patient safety, drug quality or the reliability of study results. It doesn't create an AI rule for labeling text, and it doesn't relax the labeling rules. More in is there FDA guidance on using AI to write drug labeling.
What Europe's regulator says
The European Medicines Agency's reflection paper on AI in the medicinal product lifecycle, section 2.3.5 (Product information):
AI/ML applications used for drafting, compiling, editing, translating, tailoring, or reviewing medicinal product information documents should be used under close human supervision. Given that generative language models are prone to include plausible but erroneous or incomplete output, quality review mechanisms need to be in place to ensure that all model-generated text is both factually and syntactically correct before submission for regulatory review.
What that means in practice
Our view: decide up front who owns each AI-drafted label line, record which text the tool produced, and have someone read the final high-stakes lines for meaning before approval. See how to add a read-for-meaning step and does AI drafting make wrong-word errors moot.
Sources
- LII: 21 U.S. Code § 352 - Misbranded drugs and devices
- FDA: Considerations for the Use of Artificial Intelligence To Support Regulatory Decision-Making for Drug and Biological Products
- EMA: Reflection paper on the use of AI in the medicinal product lifecycle
- eCFR: 21 CFR 207.41
- eCFR: 21 CFR 207.49
Last reviewed: October 14, 2026
More on contextual errors in FDA DailyMed labels.
Questions about wording on your own labels? Talk to us.
We scan freshly published FDA DailyMed labels every day and pull out these high-stakes wrong words.
Every Tuesday and Friday, a new case like this lands in your inbox if you subscribe.
FAQ
Who is responsible for errors in AI-drafted labeling?
Our reading: the firms that put the labeling out, such as the application holder and any repacker or relabeler that submits its own labeling. US misbranding law looks at the labeling itself, and our reading is that it makes no exception for text a tool wrote. This is general information, not legal advice.
Does the FDA's AI draft guidance cover AI used to draft labeling?
Not directly. The FDA's January 2025 draft guidance does not address AI used for drafting and writing a regulatory submission when that use doesn't affect patient safety, quality or study results.
What does EMA say about AI-drafted product information?
Its reflection paper says AI used for drafting or reviewing product information should be used under close human supervision, with quality review of all model-generated text.