Glossary · USA: FDA device and drug rules
AI-enabled device software function (AI-DSF)
Also known as: AI-DSF, AI-enabled device software function, AI-enabled medical device
In FDA terminology, an AI-enabled device software function (AI-DSF) is a device software function that implements one or more AI models to achieve its intended purpose. FDA's draft guidance of January 7, 2025 on their lifecycle management and marketing submissions describes what a submission should contain, from device description and data management to performance validation, labeling and post-market monitoring.
- USA
- AI
- Technical documentation
In one sentence
AI-DSF (FDA): a device software function that uses AI models; FDA's January 2025 lifecycle guidance for them is a draft.
Example
A sponsor of an AI-based fracture detection function describes in its 510(k) the model, the training and test data with their sources and demographics, the performance per subgroup, the user interface, the labeling with known limitations and the plan for monitoring performance after clearance.
Explained in context
This term is a station of a mini course
- The medical-device evidence chain: from intended purpose to post-market evidenceStation: Is the software a device? · Planned model changeStart the course
Context cards connect this term with others to answer one question. Also in British English and German.
- Model change approver: a functional roleWho decides whether an update of an AI model in a medical device stays inside its planned modifications — and who stops it when it doesn’t?UKDeutsch
- The medical-device evidence chain: from intended purpose to post-market evidenceHow do the parts of a medical device’s technical documentation connect — and what has to be in place before an AI function may answer from it?UKDeutsch
- Training data provenance is part of the device file, not a data-science recordWhy does the origin of training data belong in the technical documentation of an AI-enabled medical device?UKDeutsch
- Documenting the limits of a model: what the IFU of an AI-enabled device must sayHow should the limitations of an AI model be documented for the users of a medical device?UKDeutsch
- From a locked model to planned change: documenting model updates of a medical deviceHow can an AI-enabled medical device change its model without every update becoming a new regulatory question?UKDeutsch
How it applies
- Draft, not final: As of the review date the lifecycle guidance is a draft (comment period closed April 7, 2025). Cite it as draft guidance; it is not for implementation as published. The PCCP guidance for AI-enabled device software functions, by contrast, is final.
- Total product lifecycle: The draft treats design, data management, validation, labeling, cybersecurity and post-market performance monitoring as one lifecycle.
- Labeling and transparency: It recommends that users learn the model's inputs and outputs, performance including for subgroups, known limitations and how the device fits the clinical workflow — content of the IFU and the interface, sometimes in a model card format.
- Technical documentation: Keep the model version, the software version and the device version as separate identifiers and record which combination each label and IFU version describes.
AI-DSF vs. an AI system under the EU AI Act
The FDA term is a device software function that uses AI; it lives inside the device framework. The EU AI Act defines an AI system in its own right (Article 3(1)) and adds requirements to it when it is a medical device or its safety component that needs a notified body; see AI in medical devices.