Hallucination checks and source grounding

How to study this topic

Hallucination checks test whether an answer is supported by approved sources instead of model confidence or plausible wording. Read this as a banking control chapter, not as a technology marketing chapter. The useful question is whether the bank can trust, explain, control and monitor the result in source-grounded compliance AI.

The topic should stay close to banking evidence, policy ownership, customer outcome, regulatory expectation, data quality, operational process and audit trail. If the explanation drifts into generic AI language, it loses the reason this card exists.

A strong learner should be able to explain the model role, the evidence, the control owner, the human review point and the wrong-outcome risk to a business analyst, compliance analyst, credit-risk manager, developer, tester, auditor and senior risk owner.

Plain language meaning

In plain language, hallucination checks and source grounding is about turning messy banking evidence into a controlled banking answer. The bank must separate observed fact, interpretation, model output and business action.

The model may classify, retrieve, summarise, estimate or rank, but the bank decides whether to approve, refer, investigate, escalate, communicate, provision, report or remediate.

Confidence in wording is not confidence in source quality, model design, policy authority or control effectiveness. Banking AI needs proof, not just fluent output.

Where it sits in the bank

This topic normally touches risk, compliance, product, operations, legal, data, model governance, technology and internal audit. Ownership must be explicit for source, policy, model, output and action.

The relevant population includes the banking cases described by the topic, including clean cases and edge cases: missing evidence, vulnerable customers, unusual products, disputed outcomes, local regulatory differences and manual overrides.

The same AI output can be low risk in internal learning and high risk when it affects a customer, control conclusion, finance number, regulatory response or audit file. Purpose matters.

Evidence and source material

Relevant evidence includes answer claim, source citation, retrieved passage, document version, policy owner, unsupported statement, conflict marker, reviewer correction, and quality sample. These sources are not equal; official records, user-entered values, derived values, draft documents and approved policy need different trust treatment.

Timing matters because credit outcomes, policy versions, model versions, approval thresholds and customer status change. A correct answer for one date can be wrong for another date.

Evidence must be traceable to source, owner, version, date, access permission, transformation, retrieval path and limitation.

Data quality and control checks

Controls should include claim-to-source matching, citation requirement, unsupported-answer block, conflict escalation, source freshness check, human review, quality sampling, and incident logging. These controls stop weak evidence from being treated as strong evidence and stop model output moving faster than governance.

Quality means authority, completeness, business meaning, lineage, label quality, fairness, privacy, security, citation quality, output review and customer impact.

When evidence fails, the response should be known: block, limit, refer, escalate, fallback, sample, remediate or retire the use.

How AI and ML can be adopted

Useful adoption includes checking citations, flagging unsupported claims, detecting outdated references, comparing answer to source text, and monitoring correction patterns. These are support uses first; they improve search, classification, prioritisation, explanation, drafting and monitoring.

A bank should move from internal research assistance to controlled decision support and then to restricted automation only where validation, monitoring, accountability and fallback are mature.

The model role should be named with a verb: search, summarise, classify, estimate, recommend, refer, block, approve, escalate, report or communicate. Each verb has a different risk level.

Decision boundary and human judgement

The model may support judgement, but it should not erase judgement. The user must know whether the output is guidance, evidence, a draft, a score, a ranking, a referral trigger, a monitoring signal or a proposed communication.

Human review is useful only when the reviewer sees source evidence, reason codes, citations, limitations, model version, prompt context, retrieved documents and the policy rule being applied.

Overrides and corrections should be captured because they may reveal source gaps, policy ambiguity, retrieval weakness, model limitation or training needs.

Customer, compliance and conduct impact

The direct wrong outcome is a fluent answer creates false confidence in a compliance, credit, audit or customer-impact process. That is why the bank should not judge AI only by speed, test-set accuracy or user satisfaction.

A wrong model or unsupported GenAI answer can affect approval, decline, referral, complaint handling, compliance review, audit response, customer communication, collections, provisioning, capital, controls testing or regulatory reporting.

Conduct control asks what happens to the person, obligation, report or control affected by the answer. If the output creates pressure, exclusion, delay, weak disclosure or unfair treatment, it is a real banking risk.

Validation and monitoring

Validation should review concept, data, methodology, source quality, prompt design, retrieval quality, limitations, output behaviour and approved use.

Monitoring should look for drift, bad citations, outdated sources, repeated corrections, unfair outcomes, high override rates, weak explanations, user misuse, data leakage and missing audit trail.

When performance deteriorates, the response may be recalibration, retrieval tuning, source cleanup, stricter guardrails, retraining, manual review, restricted use, incident escalation or retirement.

Diagram walkthrough

The diagram follows five control steps: Draft answer, Claim check, Source match, Unsupported block, and Approved response. Read it left to right as a controlled banking flow from evidence or question through AI support and into accountable use.

Each box is a control point. A bank should be able to name the owner, source, rule, limitation and retained evidence at every step.

Bank-ready checklist

Before production use, check purpose, source authority, population, date, output role, customer impact, compliance impact and reproducibility.

Then check access control, validation, monitoring, override governance, audit evidence, fallback rules, incident response and business ownership.

If those controls are weak, the model may still produce an answer, but the bank should not treat the answer as trusted banking evidence.

Source anchors for accurate study

Basel credit-risk principles frame credit risk around a suitable credit-risk environment, sound credit granting, administration, measurement, monitoring and adequate controls.

The Basel Framework uses probability of default, loss given default and exposure at default as core credit-risk components for internal ratings based credit-risk measurement.

IFRS 9 is effective for annual periods beginning on or after 1 January 2018 and includes expected credit loss impairment requirements for financial instruments.

CECL under US GAAP estimates expected credit losses over the contractual life using historical experience, current conditions, and reasonable and supportable forecasts.

NIST AI RMF is a voluntary framework for managing risks to individuals, organisations and society from AI systems across design, development, use and evaluation.

US banking model-risk guidance expects model purpose, input quality, assumptions, limitations, validation, monitoring, governance, controls and effective challenge to be proportionate to model materiality.

Federal Reserve SR 26-2, dated 17 April 2026, supersedes SR 11-7 and SR 21-8 and attaches revised interagency guidance on model risk management for banking organisations.

The 2026 revised model-risk guidance states that generative AI and agentic AI are not within that guidance scope, while traditional statistical, quantitative and non-generative/non-agentic AI models are covered.

The EU AI Act treats AI systems used to evaluate the creditworthiness of natural persons or establish a credit score as high-risk; Union-law fraud-detection uses and prudential capital-requirement uses are carved out.

Evidence-level review

For a sample of generated compliance answers, break each into factual or normative claims. A reviewer checks whether the cited passage directly supports each claim, whether the passage was effective and authorized for the question, and whether the answer preserved conditions and exceptions. A citation to a relevant homepage is not grounding. Include tests where the corpus lacks the answer and where two versions conflict; the assistant should state uncertainty and request review. Track unsupported claims, wrong-version citations and omitted qualifications separately. A high retrieval-relevance score does not establish that the final wording is accurate or appropriate for regulatory or customer action.

Banking practice note: definition ownership

For hallucination checks and source grounding, definition ownership decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from answer claim through claim-to-source matching and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

AI adoption should make weak evidence easier to see, inconsistent treatment easier to challenge, outdated documents easier to detect and operational exceptions easier to route. It must not hide uncertainty behind confident language.

A good implementation records source, owner, version, date, transformation, retrieval path, model version, prompt context, user action, limitation, review decision and monitoring result.

Banking practice note: source authority

For hallucination checks and source grounding, source authority decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from source citation through citation requirement and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: effective-date control

For hallucination checks and source grounding, effective-date control decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from retrieved passage through unsupported-answer block and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: population design

For hallucination checks and source grounding, population design decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from document version through conflict escalation and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: policy alignment

For hallucination checks and source grounding, policy alignment decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from policy owner through source freshness check and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: model purpose

For hallucination checks and source grounding, model purpose decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from unsupported statement through human review and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: approved use

For hallucination checks and source grounding, approved use decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from conflict marker through quality sampling and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: human review

For hallucination checks and source grounding, human review decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from reviewer correction through incident logging and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: output limitation

For hallucination checks and source grounding, output limitation decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from quality sample through claim-to-source matching and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: fairness and conduct

For hallucination checks and source grounding, fairness and conduct decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from answer claim through citation requirement and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: customer harm

For hallucination checks and source grounding, customer harm decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from source citation through unsupported-answer block and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: regulatory evidence

For hallucination checks and source grounding, regulatory evidence decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from retrieved passage through conflict escalation and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: privacy and confidentiality

For hallucination checks and source grounding, privacy and confidentiality decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from document version through source freshness check and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: access control

For hallucination checks and source grounding, access control decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from policy owner through human review and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: audit trail

For hallucination checks and source grounding, audit trail decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from unsupported statement through quality sampling and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: change control

For hallucination checks and source grounding, change control decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from conflict marker through incident logging and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: monitoring thresholds

For hallucination checks and source grounding, monitoring thresholds decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from reviewer correction through claim-to-source matching and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: feedback loops

For hallucination checks and source grounding, feedback loops decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from quality sample through citation requirement and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: exception routing

For hallucination checks and source grounding, exception routing decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from answer claim through unsupported-answer block and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: incident response

For hallucination checks and source grounding, incident response decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from source citation through conflict escalation and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: committee reporting

For hallucination checks and source grounding, committee reporting decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from retrieved passage through source freshness check and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: third-party dependency

For hallucination checks and source grounding, third-party dependency decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from document version through human review and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: training and user behaviour

For hallucination checks and source grounding, training and user behaviour decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from policy owner through quality sampling and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: fallback operation

For hallucination checks and source grounding, fallback operation decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from unsupported statement through incident logging and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: retirement and redevelopment

For hallucination checks and source grounding, retirement and redevelopment decides whether the bank can trust the evidence, explain the result and defend the action. A model can produce a fluent answer or a precise score quickly, but a bank still has to prove why that output is suitable for source-grounded compliance AI.

Trace one example from conflict marker through claim-to-source matching and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: definition ownership

Trace one example from reviewer correction through citation requirement and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: source authority

Trace one example from quality sample through unsupported-answer block and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: effective-date control

Trace one example from answer claim through conflict escalation and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: population design

Trace one example from source citation through source freshness check and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: policy alignment

Trace one example from retrieved passage through human review and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: model purpose

Trace one example from document version through quality sampling and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: approved use

Trace one example from policy owner through incident logging and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: human review

Trace one example from unsupported statement through claim-to-source matching and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: output limitation

Trace one example from conflict marker through citation requirement and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: fairness and conduct

Trace one example from reviewer correction through unsupported-answer block and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: customer harm

Trace one example from quality sample through conflict escalation and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: regulatory evidence

Trace one example from answer claim through source freshness check and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: privacy and confidentiality

Trace one example from source citation through human review and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: access control

Trace one example from retrieved passage through quality sampling and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: audit trail

Trace one example from document version through incident logging and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: change control

Trace one example from policy owner through claim-to-source matching and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: monitoring thresholds

Trace one example from unsupported statement through citation requirement and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: feedback loops

Trace one example from conflict marker through unsupported-answer block and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: exception routing

Trace one example from reviewer correction through conflict escalation and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: incident response

Trace one example from quality sample through source freshness check and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: committee reporting

Trace one example from answer claim through human review and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: third-party dependency

Trace one example from source citation through quality sampling and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: training and user behaviour

Trace one example from retrieved passage through incident logging and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: fallback operation

Trace one example from document version through claim-to-source matching and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: retirement and redevelopment

Trace one example from policy owner through citation requirement and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: definition ownership

Trace one example from unsupported statement through unsupported-answer block and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: source authority

Trace one example from conflict marker through conflict escalation and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: effective-date control

Trace one example from reviewer correction through source freshness check and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: population design

Trace one example from quality sample through human review and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: policy alignment

Trace one example from answer claim through quality sampling and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: model purpose

Trace one example from source citation through incident logging and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: approved use

Trace one example from retrieved passage through claim-to-source matching and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: human review

Trace one example from document version through citation requirement and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: output limitation

Trace one example from policy owner through unsupported-answer block and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: fairness and conduct

Trace one example from unsupported statement through conflict escalation and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Banking practice note: customer harm

Trace one example from conflict marker through source freshness check and into the business use. If the team cannot trace that path without guesswork, the implementation is not mature enough for serious banking use.

Primary sources for further study

• NIST AI 600-1: Generative AI Profile
• NIST AI RMF: Measure playbook and data provenance