Marketing Submission Recommendations for a Predetermined Change Control Plan for Artificial Intelligence-Enabled Device Software Functions
FDA's final guidance (4 December 2024) on predetermined change control plans for AI-enabled device software functions — the authorization mechanism that lets a manufacturer pre-specify future model modifications in a marketing submission and implement them without a new submission. A PCCP has three components: a Description of Modifications, a Modification Protocol (the methods to develop, validate, and implement each change), and an Impact Assessment.
What this does not cover
stated in the document's own scope- Covers AI-enabled device software functions in FDA marketing submissions; it is not a general software-change policy — non-AI modifications follow FDA's existing device-modification guidances.
- A PCCP authorizes only the changes it describes: modifications outside the plan, or implemented other than per the Modification Protocol, require the normal regulatory route.
- A US pathway under §515C; other jurisdictions handle AI model change through their own mechanisms, not by recognizing an FDA PCCP.
Always verify against the current published text before relying on it for a submission or inspection.
Overview
"Marketing Submission Recommendations for a Predetermined Change Control Plan for Artificial Intelligence-Enabled Device Software Functions" is FDA's final guidance, issued 4 December 2024, on pre-authorizing change for AI-enabled devices. A predetermined change control plan is a section of a marketing submission that prospectively describes the modifications a manufacturer intends to make to an AI-enabled device software function, the methods for developing, validating, and implementing them, and their impact — so that once FDA authorizes the device with its PCCP, changes made within the plan do not require a new marketing submission. The guidance recommends a PCCP built from three components: a Description of Modifications, a Modification Protocol, and an Impact Assessment, with examples illustrating how they fit together.
Scope & applicability
AI-enabled device software functions in premarket submissions (510(k), De Novo, PMA), under the statutory authority of FD&C Act §515C. The final guidance broadened scope from machine-learning-specific to AI-enabled device software functions generally.
Legal basis & how it acquires force
A final guidance from CDRH with CBER and CDER concurrence, published 4 December 2024. Its statutory footing is section 515C of the FD&C Act — "Predetermined Change Control Plans for Devices" — added by the Food and Drug Omnibus Reform Act in December 2022, which authorizes FDA to approve or clear a PCCP within a PMA or 510(k)/De Novo pathway. The final guidance broadened the draft's machine-learning scope to AI-enabled device software functions generally.
Document structure
| Part | Covers |
|---|---|
| Background and policy | The problem of iteratively improving AI functions under a premarket system built for static devices, and the PCCP's role |
| Description of Modifications | The specific, bounded changes the manufacturer plans to make under the PCCP |
| Modification Protocol | The methods for each change: data management, re-training, performance evaluation, and update implementation |
| Impact Assessment | The benefits and risks of the planned modifications and how the protocol controls them |
| Examples | Illustrative PCCP scenarios for AI-enabled device software functions |
Key requirements
- A Description of Modifications: the specific, bounded changes the manufacturer intends to make
- A Modification Protocol: data management, re-training, performance evaluation, and update procedures for each modification
- An Impact Assessment: the benefits and risks of the planned changes and how the protocol controls them
- Modifications implemented per the authorized PCCP require no new marketing submission; changes outside it follow normal change rules
Implementation tips
- Write the Description of Modifications narrowly enough to be verifiable — a PCCP is authorization for specified changes, not a licence for open-ended model evolution
- Version the PCCP with the device's change-control system so post-authorization drift between plan and practice is caught internally, not by FDA
Revision notes
Final 4 December 2024, implementing the PCCP authority Congress added as FD&C Act §515C in December 2022; the final version covers AI-enabled functions generally, not only machine learning.
Where this control fails
live FDA enforcementLive FDA recalls SPEQ maps to this standard’s topics — a SPEQ interpretation, not an FDA classification.
International alignment
The guidance implements FD&C Act §515C for the device class where iterative change is the point: AI-enabled software functions. It presumes the vocabulary of the IMDRF SaMD framework (N10/N12) and the engineering discipline of IEC 62304, and its change-control philosophy extends the device QMS: modifications within the authorized plan are executed under the manufacturer's quality system rather than through new submissions. Changes outside the PCCP fall back to FDA's ordinary modification rules for the pathway.
FDA PCCP for AI-Enabled DSF (2024): frequently asked questions
Quick answers to common questions about FDA PCCP for AI-Enabled DSF (2024).
What is a predetermined change control plan?
A section of a marketing submission that prospectively specifies the modifications a manufacturer intends to make to an AI-enabled device software function, the methods that will develop and validate each change, and their impact. Once authorized, changes made consistent with the PCCP do not require a new marketing submission.
What are the three components of a PCCP?
A Description of Modifications (the specific planned changes), a Modification Protocol (the data management, re-training, performance-evaluation, and implementation methods for each change), and an Impact Assessment (the benefits and risks of the changes and how they are controlled).
What legal authority does a PCCP rest on?
Section 515C of the FD&C Act, added by the Food and Drug Omnibus Reform Act of 2022, which authorizes FDA to clear or approve predetermined change control plans in 510(k), De Novo, and PMA submissions. The December 2024 guidance describes how FDA recommends using that authority for AI-enabled device software functions.
Did the final guidance change scope from the draft?
Yes. The 2023 draft addressed machine-learning-enabled device software functions; the final guidance of 4 December 2024 covers AI-enabled device software functions generally.
This standard in practice
Recall domain is a SPEQ mapping of this standard’s topics, not an FDA classification.