UK Customs' New Standard Says AI-Prepared Declarations Should Be Disclosed. Here's What That Means.

A new voluntary standard for UK customs intermediaries was launched on 2 June 2026. Published by the British Standards Institution and sponsored by HMRC, PAS 41201:2026 sets out good practice for the preparation and submission of customs declarations. It covers due diligence, staff training, complaints handling, auditing and transparency.

The transparency section contains an important provision for intermediaries using AI. Under Section 9.1, which deals with service offerings, HMRC's guidance states plainly: "If you are using AI to assist with declaration completion, this should be disclosed to the principal."

That is not a statutory obligation. PAS 41201 is a voluntary standard, not legislation. But it represents the first time best practice for customs intermediaries has been formally set out in the UK, and it reflects where the regulatory direction of travel is heading. Intermediaries who want to align with the standard, or who are preparing for the certification scheme that HMRC is now developing alongside it, need to take the AI disclosure provision seriously.

PAS 41201 Explained

PAS 41201 was developed by HMRC in partnership with BSI and industry representatives, following a call for evidence in 2022 and a public consultation in 2023. It is designed to raise quality and consistency across the customs intermediary sector without introducing mandatory regulation.

The standard applies at organisational level, not individual. It uses normative language ("shall") for requirements that must be met to claim compliance, and informative language for recommended best practice. Intermediaries can assess themselves against it now. A separate UKAS-accredited certification scheme is being developed, and once available, certification will also be voluntary.

The AI disclosure provision sits within Section 9.1, the transparency and service offering section. It is not a standalone AI regulation. It is part of a broader requirement for intermediaries to give their principals a clear picture of how their declarations are being prepared.

What Has Changed

The customs industry has moved quickly to adopt AI tools. Document extraction, classification support, validation checking and declaration pre-population are all areas where AI is now being used, in some cases by intermediaries who have not yet thought carefully about what that means for their client relationships or their internal governance.

The appeal is obvious. One UK freight forwarder described a single consolidated invoice taking five hours to process manually, with staff retyping origin details, HS codes, descriptions and values from vendor paperwork into Excel before a declaration could begin. When AI can reduce that burden substantially, the operational case is straightforward.

What PAS 41201 adds is a governance framework around that adoption. The standard does not say AI is problematic. What it does is connect AI use to the transparency obligations intermediaries owe their clients: if you use it, say so. Used with clear processes, regular review and human oversight, AI has a legitimate role in declaration preparation.

The AI disclosure provision in Section 9.1 is part of a wider transparency requirement. Intermediaries are expected to be clear about the services they provide, how declarations are prepared, whether AI is used, and what clients can expect in the event of a customs audit after the relationship ends. Disclosing AI use is one element of that picture, not a separate or uniquely onerous obligation.

The more operationally significant provisions are in Clause 4.2, which sets out the standard's auditing requirements.

Why It Matters: The Auditing Requirement Is the Harder Problem

The auditing provisions of PAS 41201 are where the operational lift is greatest, and where AI governance becomes most consequential.

Under Clause 4.2, intermediaries who align with the standard are expected to conduct:

• An annual audit of systems and processes
• Quarterly quality control audits of submitted declarations, covering a minimum 5% sample up to a maximum of 25,000 declarations
• Documented findings, with systemic issues identified and rectified

These are not AI-specific requirements. They apply to all declaration preparation. But they change the calculus for intermediaries using AI significantly.

If AI is handling document extraction and classification support across a high volume of declarations, a 5% quarterly sample audit is a meaningful governance mechanism. It creates a documented record of declaration quality. It requires intermediaries to identify and correct systemic errors. And it means that AI-assisted processes need to be auditable in a practical sense: reviewers need to be able to understand what happened in a given declaration, why data was captured as it was, and where human review occurred.

That requires more than generating an accurate output. It requires processes and records that allow the result to be reviewed.

Three practical steps for intermediaries using AI

These are not three separately prescribed AI requirements of PAS 41201. They are practical steps that follow logically from aligning with the standard's transparency and auditing provisions.

1. Disclose where AI is used in declaration preparation. Section 9.1 requires intermediaries to be transparent about their service offerings, what principals can expect during declaration submission, and what happens in the event of a customs audit after the relationship ends. The AI disclosure provision sits within that section: if AI assists with declaration completion, the principal should be told. Clients choosing an intermediary may ask for evidence of how specific requirements are met.

2. Maintain records that support audit and review. Clause 4.2 requires quarterly declaration audits. For AI-assisted declarations, sensible practice involves maintaining records that link outputs to source documents and identify where human review and correction occurred. The goal is having enough of a record to conduct a meaningful quality control review and demonstrate what happened when things go wrong. Clause 4.2 does not prescribe specific technical requirements for how AI outputs are recorded.

3. Establish human oversight and exception handling. In practice, that means defining which decisions AI can support, which require human approval, and how uncertain or flagged outputs are escalated. This is good operational design regardless of the standard, and it aligns with the broader intent of PAS 41201's transparency and auditing provisions.

None of these steps requires replacing existing systems. They require being deliberate about how AI sits within a declaration workflow, and being able to demonstrate that deliberateness to a client or auditor.

What Good Looks Like in Practice

The gap between using AI in a customs workflow and using it in a way that supports PAS 41201 alignment is largely a governance gap, not a technology gap.

Good AI-assisted declaration preparation has a recognisable shape. AI handles document extraction and classification support, applying the intermediary's business rules and knowledge of the relevant tariff. Outputs are reviewed by qualified operators before submission. Uncertain or flagged items are escalated rather than passed through silently. Source documents are retained alongside declaration records so that any output can be traced back to its origin. Human operators retain responsibility for review and submission.

That last point matters. AI can reduce the manual effort involved in preparing a declaration substantially, but the intermediary remains responsible for fulfilling its obligations, while legal liability also depends on the customs representation arrangements. The standard does not change that responsibility; it asks intermediaries to be transparent about how they are exercising it.

Where AI governance commonly falls short

Conversations with customs operations teams reveal consistent patterns. Data validation is frequently underweighted relative to data extraction. It is not enough for AI to pull figures from a document; the outputs need to be checked for internal consistency before submission. An operations team we work with highlighted that certain field relationships, such as weight data, can cause declaration acceptance issues at CDS if they are not validated correctly. That kind of domain-specific checking is part of what separates AI that reduces workload from AI that transfers risk.

Classification maintenance is another area where implementations drift. Commodity codes change. Equipment classifications go stale. Codes used under one tariff schedule require verification against another. One logistics operation we work with needed a live classification catalogue connected to their asset management system, updated as rulings change, rather than a one-time exercise that becomes unreliable over time. The quarterly audit requirement in Clause 4.2 creates a structural incentive to keep classification data current: errors found in sampling need to be corrected and documented.

Adoption consistency is a practical consideration that PAS 41201 does not address directly. One UK forwarder achieved strong AI system adoption at one office, while other branches were still in early stages months into deployment. Inconsistent use can make disclosure and process controls harder to manage, and is exactly the kind of variation that a quarterly audit is designed to surface.

Before deployment, the integration question also needs an honest answer. One international forwarder found that compatibility with their existing freight management system required development work to confirm before user acceptance testing could proceed. Integration determines whether AI delivers end-to-end value or creates a new manual handoff point.

The Direction of Travel

PAS 41201 is a voluntary standard. It does not automatically become a legal requirement, and the AI disclosure provision in Section 9.1 does not impose new statutory obligations on intermediaries.

What it does is establish a benchmark. A separate HMRC consultation on mandatory registration for customs intermediaries opened on 22 June 2026 and closed on 21 September 2026. That consultation is a distinct policy proposal, covering the scope of registration, applicable requirements and enforcement approach. Its outcome has not yet been announced as of the date of publication. What it signals, alongside PAS 41201, is that the UK government is actively shaping what it expects from the customs intermediary sector over the next several years.

Intermediaries who align with PAS 41201 now are building governance infrastructure that could prove valuable if mandatory requirements are introduced. The auditing provisions, the transparency requirements and the human oversight that good AI governance needs are not burdens invented for their own sake. They are the conditions under which AI-assisted declaration preparation can be operated responsibly, at scale, with accountability to clients and to HMRC.

For customs intermediaries using AI, PAS 41201 provides a useful benchmark. Disclosing where AI is used is a starting point. The bigger operational challenge is ensuring that declaration preparation remains accurate, traceable and subject to appropriate human oversight.

As AI takes on more of the work involved in preparing customs declarations, those controls will become increasingly important.

The intermediaries that put them in place now will be better positioned to demonstrate the quality of their operations to customers, auditors and regulators.

Frequently Asked Questions

Does PAS 41201:2026 require customs intermediaries to disclose AI use? PAS 41201 is a voluntary standard, not legislation. Section 9.1 states that if AI is used to assist with declaration completion, this should be disclosed to the principal. Intermediaries who align with the standard are expected to make that disclosure as part of their broader transparency obligations around service offerings.

Is PAS 41201 mandatory for UK customs intermediaries? No. PAS 41201 is a voluntary standard, sponsored by HMRC and published by BSI. It sets out good practice for the preparation and submission of customs declarations. A separate HMRC consultation on mandatory registration for customs intermediaries closed in September 2026, but its outcome has not yet been announced. PAS 41201 itself does not impose statutory obligations.

What auditing does PAS 41201 require? Under Clause 4.2, intermediaries who align with the standard are expected to conduct an annual audit of systems and processes, plus quarterly quality control audits of submitted declarations covering a minimum 5% sample up to a maximum of 25,000 declarations. Findings should be documented and systemic issues identified and rectified.

What does "auditable AI" mean in a customs context? It means maintaining records that allow a declaration to be reviewed: linking AI-generated outputs to source documents, identifying where human review and correction occurred, and being able to demonstrate what happened if a client or auditor asks. Clause 4.2 does not prescribe specific technical requirements for how AI outputs are recorded, but the quarterly audit process creates a practical need for that level of traceability.

How does PAS 41201 differ from AEO status? Authorised Economic Operator status is an internationally recognised authorisation focused on supply chain security, covering areas such as financial standing, record keeping and procedures for controlled goods. PAS 41201 is designed to be customer-facing, focusing on service quality, transparency and professional development. The two are complementary but distinct. PAS 41201 does not replace AEO or any other existing authorisation.

Adrian Smith is Co-Founder and CEO of Ripple, which designs, builds and runs enterprise-grade AI solutions for logistics and supply chain operations. Ripple's customs classification and clearance agents are in production with freight forwarders and logistics providers across the UK and internationally.‍