Future & research4 min read

Data quality and ethics

AI learns from data — the quality, consent and governance of that data therefore decide whether AI is safe, fair and ethical.

“Garbage in, garbage out” is the law of machine learning. Models trained on noisy, incomplete or biased data make noisy, incomplete or biased decisions. The medical AI community therefore treats data curation as a core engineering discipline.

Ethics is the second half of the story. Imaging data is deeply personal; its use demands informed consent, strict anonymisation, protection against breaches and governance that puts patients first. Transparent data policies are what make AI acceptable in medicine.

Foundations of responsible AI

  • Curated, labelled, clinically validated training data
  • Informed consent and transparency for patients
  • Strong anonymisation and security
  • Independent governance and audits

In brief

Model quality depends directly on data quality.
Consent and anonymisation are non-negotiable for patient data.
Independent governance builds public trust.

Keep learning

Bias and fairness in AIAI HubSafety and data privacyAI HubClinical validation and evidenceAI Hub