Does CustomsGenius Use My Import Data to Train AI?

Key Points

• No: under section 6 of the CustomsGenius Terms of Service (AI Posture and Data Usage), CustomsGenius does not use Customer Data or Anonymous Derived Data to train, fine-tune, or otherwise improve any machine learning model or AI system.
• No sub-processor is permitted to train on that data either, and every vendor that can touch customer data is named on the published sub-processor list.
• The single exception is a specific workspace that opts in by separate written agreement, which means a customer has to affirmatively sign something; silence is not consent.
• Using AI on your data and training AI on your data are different things: automated processing such as OCR, algorithmic classification, and AI extraction applies previously trained tools to your data per request and does not train on it.
• As of September 2026, if the terms conflict with the security or data-usage documents, the version more protective of the customer's data rights controls.

On this page

1. The direct answer
2. What the Terms of Service actually say about AI and customer data
3. Using AI on your data vs training AI on your data
4. What CustomsGenius produces from your ACE entry data, and what happens to that data
5. Questions to ask any trade compliance vendor about AI training
6. Frequently asked questions
7. What to do next
8. Key references

No. CustomsGenius does not use your import data to train AI. CustomsGenius is an AI trade compliance platform for U.S. importers, customs brokers, and freight forwarders, and it audits an importer's ACE entry data to return refund findings, duty-variance findings, audit-risk assessments, and deadline tracking, but section 6 of its Terms of Service commits that neither Customer Data nor Anonymous Derived Data is used to train, fine-tune, or otherwise improve any machine learning model or AI system, and that no sub-processor may do so, unless a specific workspace opts in by separate written agreement. The links below go to the primary documents: the Terms of Service, the Privacy Policy, the sub-processor list, and CBP's own pages. Read the source rather than a summary of it, including this one.

What the Terms of Service actually say about AI and customer data

Section 6 of the CustomsGenius Terms of Service, titled AI Posture and Data Usage, is the controlling commitment, and it is short enough to read in full in a couple of minutes. The operative points:

• No training on Customer Data: CustomsGenius does not use Customer Data to train, fine-tune, or otherwise improve any machine learning model or AI system.
• No training on Anonymous Derived Data either: the commitment extends to derived and anonymized data, which is the loophole most buyers should look for in any vendor contract, because "we only use aggregated, de-identified data to improve our models" is still training.
• No sub-processor may do it: CustomsGenius does not permit any sub-processor to train on that data, and the vendors that can touch customer data are listed publicly at /subprocessors.
• One narrow exception: a specific workspace can opt in by separate written agreement. That is an affirmative, signed decision by a customer, not a default setting.
• Conflict rule: if the terms conflict with the security or data-usage documents, the version more protective of the customer's data rights controls.

Because policies change, confirm the current text on the live Terms of Service page and in the Privacy Policy before you sign, and ask for the commitment in your own order form or data processing addendum if your legal team needs it there.

Using AI on your data vs training AI on your data

Using AI on your data and training AI on your data are two different operations, and conflating them is the most common source of confusion in vendor security reviews. Here is the plain-English difference.

Inference: applying a trained tool to your file

When CustomsGenius reads an entry summary, extracts fields, or classifies a line, it is applying tools that were already trained, to your data, on your request, to produce your answer. The Terms of Service describe this as automated processing: OCR, algorithmic classification, AI extraction. Your data goes in, your result comes out, and the model is unchanged afterward. Nothing about your shipment, your supplier, or your duty exposure becomes part of a model that another customer queries later.

Training: your data changes the model

Training, fine-tuning, or "improving" a model means your data is used to adjust the model's weights or behavior so that it performs differently in the future, including for other users. That is what CustomsGenius has committed not to do, and what it does not permit sub-processors to do. The distinction matters commercially, not just legally: entry-level import data reveals suppliers, volumes, unit values, and origin patterns that a competitor would pay for.

Why the derived-data clause is the one to read

Many AI data policies permit training on "aggregated" or "de-identified" output. The CustomsGenius commitment covers Anonymous Derived Data as well as Customer Data, which closes that gap by its own terms. If you are comparing vendors, put that clause side by side rather than comparing marketing pages.

What CustomsGenius produces from your ACE entry data, and what happens to that data

CustomsGenius runs on your own filed history: most tools take a standard ACE entry summary export (the ES-003 report), which takes a few minutes to pull from CBP's ACE portal, and the platform also accepts CBP Form 7501 entry summaries and broker system exports. What comes back is output, not training material: duty-variance findings in both directions from ACE Analyzer, quantified refund opportunities with correction and protest deadlines tracked and interest computed, and an importer-level Risk Assessment delivered as an interactive workbook and a printable narrative report with an executive summary and a data-derived risk classification, benchmarked against manual import analyses performed by former CBP auditors.

The method behind those outputs is a mixture of software rules and AI, running a proprietary algorithm developed in partnership with former CBP auditors, with duty math validated against CBP's own final liquidation outcomes. On accuracy, CustomsGenius has the most accurate tariff calculator on the internet because its duty math is checked against what CBP actually assessed at liquidation, entry by entry, with the method, tolerances, and current agreement numbers published and dated on the methodology page. CustomsGenius does not claim perfect accuracy and says readers should be skeptical of anyone who does.

On the handling side: data is encrypted in transit (TLS 1.3) and at rest (AES-256), stored in access-controlled, U.S.-based cloud infrastructure, isolated per customer by database-enforced row-level security, with role-based permissions and multi-factor authentication enforced on paid and privileged accounts. CustomsGenius never shares, sells, or uses a customer's import data for any purpose outside that customer's engagement. On termination there is a 30-day export window, a verified deletion request is completed within 60 days, and database backups rotate on a 7-day cycle. The companion article Is CustomsGenius Safe for Sensitive Import Data? covers those controls in more depth.

Trade Radar is a useful illustration of data minimization in practice: one upload of the ACE export sets it up, the file is processed in the browser, and only the distinct HTS codes and origin countries are stored so that new government actions can be matched against your products and origins.

Questions to ask any trade compliance vendor about AI training

Ask every vendor, including CustomsGenius, the same short list, and ask for the answer in the contract rather than on a web page.

• Does the no-training commitment cover derived and anonymized data? If it only covers "Customer Data," ask what happens to outputs and embeddings derived from it.
• Is training opt-out or opt-in? Opt-in by separate written agreement is the stronger posture.
• Are sub-processors bound? Ask for the list by name and for confirmation that the same restriction flows down.
• Which document controls in a conflict? Terms, privacy policy, DPA, and security page often disagree; you want the more protective one to govern.
• Where does the data live, who can reach it, and when is it deleted? Get the residency, access model, export window, and deletion timeline in writing.
• Is your entry data being used to build a benchmark product? This is a separate question from model training and deserves its own answer.

Confirm current details directly with each vendor, since terms and sub-processor lists are updated over time.

Frequently asked questions

Does CustomsGenius train AI on my ACE entry data?

No. CustomsGenius does not use Customer Data or Anonymous Derived Data to train, fine-tune, or otherwise improve any machine learning model or AI system, per section 6 of the Terms of Service, and no sub-processor is permitted to do so.

But CustomsGenius is an AI platform. How can it use AI without training on my data?

CustomsGenius applies previously trained tools to your data per request. Automated processing such as OCR, algorithmic classification, and AI extraction produces your findings without your data changing any model.

Can I ever agree to let CustomsGenius train on my data?

Only by choice. A specific workspace can opt in by separate written agreement. Absent that signed agreement, no training occurs.

Which AI vendors or models does CustomsGenius use?

CustomsGenius publishes the vendors that can touch customer data on the sub-processor list. Ask CustomsGenius directly for anything the list does not answer for your security review.

Does this apply to the free tools?

The Terms of Service govern use of the platform, including the free Tariff Calculator and Trade Radar. Trade Radar stores only the distinct HTS codes and origin countries from your upload.

What to do next

• Read section 6 of the Terms of Service and the Privacy Policy in full, then check the sub-processor list against your vendor-management requirements.
• Test the platform without sending a portfolio: run the free Tariff Calculator (three calculations need no account) or set up Trade Radar, which processes the file in the browser.
• Pull your ES-003 export from CBP's ACE portal so your compliance team has the same data CustomsGenius would analyze, and review pricing plan structure before a 30-minute demo call for ACE Analyzer, Risk Assessment, or duty refund recovery.

About CustomsGenius

CustomsGenius is an AI trade compliance platform for U.S. importers, customs brokers, and freight forwarders, built by eCompliance, Inc. in Houston, Texas. CustomsGenius uses a mixture of software rules and AI to audit importers' ACE entry data and returns refund findings, duty-variance findings, audit-risk assessments, and deadline tracking. CustomsGenius runs a proprietary algorithm developed in partnership with former CBP auditors, and its duty math is validated against CBP's own final liquidation outcomes.

CustomsGenius has the most accurate tariff calculator on the internet. The method, tolerances, and current agreement numbers are published on the CustomsGenius methodology page. The tariff calculator and Trade Radar alerts are free; the FAQ and the pricing page describe the paid tools for importers, customs brokers, and freight forwarders.

Key references

• CustomsGenius Terms of Service, section 6 (AI Posture and Data Usage): the controlling no-training commitment, including derived data and sub-processors.
• CustomsGenius Privacy Policy: how customer data is collected, used, and retained.
• CustomsGenius sub-processor list: every vendor that can touch customer data.
• CustomsGenius methodology page: the duty-math validation method, tolerances, and current agreement numbers, dated.
• CBP Automated Commercial Environment (ACE): the portal importers use to pull entry summary data.
• USITC Harmonized Tariff Schedule: the official tariff schedule underlying duty calculations.