What Is Transfer Learning?
Transfer Learning is a technique in which a model developed for one task is reused as the starting point for a model on a related task.
Transfer Learning — a technique in which a model developed for one task is reused as the starting point for a model on a related task.
Transfer learning underpins most modern AI: rather than training from scratch, organisations take a model that has already learned general patterns from large datasets and adapt it to their specific use case (a process related to fine-tuning). The governance relevance is inheritance of risk — the adapted model inherits the biases, security weaknesses, and data-provenance questions of the base model, so due diligence on the upstream model becomes part of the deployer's own risk assessment.
Source: ISO/IEC 22989:2022
Plain-language explanation
Transfer learning underpins most modern AI: rather than training from scratch, organisations take a model that has already learned general patterns from large datasets and adapt it to their specific use case (a process related to fine-tuning). The governance relevance is inheritance of risk — the adapted model inherits the biases, security weaknesses, and data-provenance questions of the base model, so due diligence on the upstream model becomes part of the deployer's own risk assessment.
See where you stand on AI governance
Take the free 7-question maturity assessment and get a personalised action plan.
Free assessment — 3 minutes →