Category: AI in insurance · Reviewed by Al Jabbar, Broker · Specialist Risks · Last reviewed 2026-06-10
Explainable AI (XAI) in insurance is the set of techniques and governance practices used to make the decisions of artificial-intelligence and machine-learning models understandable to those affected by them and to those accountable for them. In the United Kingdom market the expectation is anchored in UK GDPR Article 22, the FCA Consumer Duty, the ICO / Alan Turing Institute guidance on explaining AI decisions, and the supervisory direction in DP5/22 and FS2/23.
intrinsic interpretability of model architectures (linear models, decision rules, monotonic GBMs); and
post-hoc explanation techniques applied to opaque models, including SHAP (Shapley additive explanations), LIME (local interpretable model-agnostic explanations), partial-dependence plots, integrated gradients (for neural networks) and global surrogate models.
For UK insurers the practical question is not whether a model is “explainable in the abstract” but whether the firm can deliver appropriate explanations to consumers, distributors, regulators and its own governance bodies.
Legal / Regulatory basis
UK GDPR Article 22 restricts solely automated decisions that produce legal or similarly significant effects, requiring meaningful safeguards including the right to obtain human intervention and to contest the decision.
Data Protection Act 2018 and UK GDPR Article 13–15 transparency rights — “meaningful information about the logic involved” in automated processing.
ICO and The Alan Turing Institute, Explaining decisions made with AI (May 2020) — the canonical UK guidance, organised around six types of explanation: rationale, responsibility, data, fairness, safety and impact.
ICO Generative AI consultation responses (2024) — extending these expectations to generative models.
FCA Consumer Duty (PS22/9) — the consumer-understanding cross-cutting rule means a firm must be able to communicate, in plain language, why a price was reached or a claim was treated in a particular way.
FCA & PRA DP5/22 and FS2/23 — explainability is repeatedly cited as a governance pillar.
EIOPA AI Governance Principles (June 2021) — explainability is one of the six principles.
EU AI Act (Regulation (EU) 2024/1689) — for UK insurers with EU operations, high-risk AI systems (which include insurance pricing of life and health) are subject to transparency obligations.
How it works in practice
Different audiences require different levels of explanation:
Consumer-facing explanations — typically short, plain-English statements consistent with the Consumer Duty; in motor pricing, for example, a list of headline factors influencing the price.
Distributor / broker explanations — slightly richer, suitable for fact-finding under ICOBS 5.
Internal validation — detailed SHAP / LIME outputs; global feature importance; partial-dependence and individual conditional expectation plots; counterfactual examples.
Regulator / audit — full model-risk file: training data, methodology, validation, fairness slicing, ongoing monitoring.
Right-to-object / Article 22 — documented procedures for human review of contested automated decisions.
The ICO / Turing six-explanation framework is now embedded in many UK firms’ model-risk policies.
Surrogate models — fitting a simple model to mimic an opaque model’s outputs.
Counterfactual explanations — “what minimum change would have produced a different outcome?”.
Explanation faithfulness testing — checking that post-hoc explanations actually reflect the underlying model.
Generative AI explainability — chain-of-thought summaries, retrieval citations and provenance metadata for LLM outputs (see Large language model (LLM) insurance).
Example
A UK home insurer rebuilds its pricing on a GBM. The firm’s XAI policy adopts the ICO / Turing six-explanation framework. Customers receive a plain-English summary of the top three drivers of their renewal price; brokers can access a more detailed SHAP visualisation in the portal; and the model-risk file includes faithfulness checks and counterfactual examples. The procedure for human review under UK GDPR Article 22 is documented in the firm’s privacy notice and on the policy schedule.
ICO and The Alan Turing Institute, Explaining decisions made with AI, May 2020. https://ico.org.uk/for-organisations/guide-to-data-protection/key-dp-themes/explaining-decisions-made-with-artificial-intelligence/
ICO, Generative AI consultation responses, 2024. https://ico.org.uk
UK GDPR Articles 13–15 and 22; Data Protection Act 2018, https://www.legislation.gov.uk
FCA, PS22/9 — Consumer Duty, July 2022.
FCA & PRA, DP5/22 and FS2/23 — AI and Machine Learning, October 2022 / 2023.
EIOPA, Artificial Intelligence Governance Principles, June 2021.
Regulation (EU) 2024/1689 (EU AI Act).
Lundberg, S. & Lee, S., “A Unified Approach to Interpreting Model Predictions” (SHAP), NeurIPS, 2017.
Ribeiro, M. et al., “Why Should I Trust You? Explaining the Predictions of Any Classifier” (LIME), KDD, 2016.
This entry is part of the Apex Insurance Wiki. Last reviewed by Matt Bartlett on 2026-06-10. Next review: 2026-12-10.
Apex Insurance Brokers Limited. Authorised and regulated by the Financial Conduct Authority, FRN 724952. Registered in England and Wales, Companies House 07014570. This entry provides general information about UK insurance concepts and is not regulated advice. Consult your insurance broker on your specific position.
Our service promise. We acknowledge every quote request the same working day. For straightforward risks, indicative terms typically follow within five working days. Complex risks — higher-risk buildings, cladding, mid-term proposals requiring fresh underwriting — may take longer; we’ll send you a progress note by the end of the fifth working day in those cases.