An FDA warning letter to a pharmaceutical manufacturer cites non-compliant use of AI use in manufacturer quality systems. It is also a warning about using AI in medical device manufacturing outside the typical algorithm and machine learning applications.

The FDA sent a warning letter to Purolea Cosmetics Lab on April 2, 2026. The part about AI is new and important as it addresses using AI in a Current Good Manufacturing Practice (or cGMP). The warning letter link and text extract are included at the end of this post for reference.

What the FDA Said About AI

The company told FDA inspectors that they used “AI agents” to write:

• product specifications,
• step-by-step procedures, and
• master production and control records.

Why? They wanted AI to help achieve compliance. The FDA’s response was clear:

“If you use AI as an aid in document creation, you must review the AI generated documents to ensure they were accurate and actually compliant with CGMP. Your failure to do so is a violation of 21 CFR 211.22(c). Overreliance on artificial intelligence for your drug manufacturing operations was also documented during the inspection.”

The company also said the AI never told them they needed to do process validation (a required step under rule 21 CFR 211.100). The FDA said that excuse does not work. The rules still apply fully, and AI cannot take the blame or replace human checks.

This letter does not ban AI. It is a warning about using AI inappropriately as a tool. The company’s quality unit must still review, approve, and make sure everything is in compliance.

How This Applies to Medical Device Manufacturers

The letter does not mention medical devices, but the main idea carries over for AI use in any manufacturer quality systems.

• Quality system rules: Device makers must still follow rules about management, design controls, production, and process validation. If AI writes design plans, risk reports, manufacturing steps, or inspection records, people must review and approve them. Overreliance without checks could be treated as a violation, just like in the drug letter.
• When the device itself uses AI: This has been the historical focus for medical device AI use. The FDA has separate guidance on how to manage the full life cycle of these AI tools. You still need human oversight, risk checks, and plans for updates. The warning letter is a reminder that you cannot hand off final responsibility to AI here, either.
• AI used only in manufacturing: Tools for automated checks, predictive maintenance, or document writing in a device factory must still meet the same quality system rules. The letter shows the FDA will now look closely at AI use during inspections.

Bottom line for device makers: Use the warning letter as a heads-up. Put appropriate rules in your quality system for how you handle AI. This is also important if you use AI for design files, problem reports, or automatic quality tests.

Bigger Picture and What It Means

• AI is a helper, not the boss: Both drug and device rules say the manufacturer (not the AI) is fully responsible. This warning letter makes that point very clear.
• Matches FDA’s other AI rules: It fits with FDA guidance on AI in manufacturing, decision-making, and device software. Those documents all stress risk management, good data, model testing, and human control.
• Future inspections: The FDA will probably ask more questions about AI during inspections. Companies should expect inspectors to check: “Who looked over and signed off on the AI output?”

What Drug and Device Companies Should Do for Their Quality Systems (if you haven’t already)

1. List every place you use AI in your regulated work.
2. Update your procedures so a real person must always review and approve AI results.
3. Test and control your quality system AI tools based on how much risk they carry.
4. Train your quality team on what AI can and cannot do, and why the company stays responsible.
5. Add AI rules to your overall quality system and check them during internal audits.

This one warning letter, even though it went to a small lab, sets an important example for the whole industry. It does not stop companies from using AI. It just reinforces to use AI in a safe, secure, and compliant way, including AI use in manufacturer quality systems.

Warning Letter Link and Extract

Extract from FDA warning letter: https://www.fda.gov/inspections-compliance-enforcement-and-criminal-investigations/warning-letters/purolea-cosmetics-lab-722591-04022026

3. Your firm’s quality control unit failed to exercise its responsibility to ensure drug products manufactured are in compliance with CGMP and meet established specifications for identity, strength, quality, and purity (21 CFR 211.22).

Your quality unit (QU) did not adequately oversee your drug manufacturing operations. For example, your QU failed to ensure:

• Procedures were established or followed (21 CFR 211.22(d))
• Batch records were reviewed before drug product release (21 CFR 211.22(a))
• Adequate production and process controls were established (21 CFR 211.100(a))

Your firm’s quality systems are inadequate. See FDA’s guidance document, Quality Systems Approach to Pharmaceutical CGMP Regulations, for help implementing quality systems and risk management approaches to meet the requirements of CGMP regulations 21 CFR, parts 210 and 211 at https://www.fda.gov/media/71023/download.

Inappropriate Use of Artificial Intelligence in Pharmaceutical Manufacturing.

During the FDA inspection of your drug manufacturing facility, you stated to FDA investigators that you utilized artificial intelligence (AI) agents (b)(4) to help your firm comply with FDA regulations. Specifically, you used AI to create drug product specifications, procedures, and master production or control records to be in compliance with FDA requirements.

If you use AI as an aid in document creation, you must review the AI generated documents to ensure they were accurate and actually compliant with CGMP. Your failure to do so is a violation of 21 CFR 211.22(c). Overreliance on artificial intelligence for your drug manufacturing operations was also documented during the inspection. For example, the FDA investigators found that you had not conducted process validation prior to distribution of your drug products, as required under 21 CFR 211.100, and informed you as such. You replied that you were not aware of the legal requirement, as the AI agent you used (b)(4), never told you it was required.