Introduction
Artificial intelligence can help compliance teams work faster and focus on real risk. It can also make mistakes. Nex Global LLC, trading as Nex Systems (Nex), believes customers, regulators and the people whose names are screened deserve to know how and where Nex uses AI, and what its limits are.
This AI Use and Transparency Statement (the Statement) explains the AI features in Nex services, the principles Nex follows, the data those features use, and the responsibilities that stay with the customer. It forms part of, and is incorporated by reference into, the Nex Systems Terms of Service (the Terms). It should be read with the Privacy Notice, the Data Processing Agreement and the Acceptable Use Policy.
In short: Nex AI supports human judgement. It does not replace it.
1. Where and How We Use AI
1.1 AI-assisted features. Nex uses AI, including machine learning and large language models, in the following kinds of features. In the product, these features are presented under the name Elodie AI. Not every feature is available on every plan.
| Feature | What AI does | What the user does |
|---|---|---|
| Adverse-media filtering | Filters and ranks news and open-source results by relevance to the person or entity screened, and removes likely duplicates and irrelevant articles | Reviews the remaining articles and decides whether they are relevant and adverse |
| Name matching assistance | Helps compare names across scripts, spellings and transliterations, and explains why a potential match was raised | Reviews each potential match and confirms or discounts it |
| Summaries | Produces short summaries of profiles, articles or case history | Reads the source material and relies on it, not on the summary alone |
| Risk-scoring suggestions | Suggests risk factors or a risk level based on the customer's own configured methodology | Accepts, changes or rejects the suggestion and records the reason |
| Compliance assistant | Answers questions about using the services and general compliance concepts | Treats answers as general information, not legal advice, and checks primary sources |
1.2 Labelling. AI-generated content in the services is identified as such, for example by a label or icon, so users know when they are looking at AI output.
1.3 Non-AI functions. Sanctions list collection, list updates and core exact and fuzzy matching run on deterministic rules and are not generative AI.
2. Our AI Principles
2.1 Nex designs, builds and operates its AI features according to these principles:
| Principle | What it means at Nex |
|---|---|
| Human in control | AI makes suggestions. A trained person at the customer reviews them and makes every compliance decision. |
| Transparency | We tell users when content is AI-generated and, where practical, show the sources and reasons behind it. |
| Explainability | Features that raise or rank potential matches show the factors that led to the result, so users can test it. |
| Fairness | We test for and work to reduce unfair bias, for example in how names from different languages and cultures are matched. |
| Privacy and security | AI features follow the same security controls, data minimisation and confidentiality rules as the rest of the services. |
| Reliability | We test AI features before release and monitor them after release, and we can switch a feature off if it does not perform as expected. |
| Accountability | Named people at Nex are responsible for each AI feature, and for responding to feedback and incidents. |
3. What Our AI Does Not Do
3.1 Nex AI features do not:
- (a) make final decisions to accept, reject, block, exit or report a customer, transaction or person;
- (b) make decisions that produce legal or similarly significant effects on individuals without human review;
- (c) file, draft for filing, or submit any suspicious activity report or other report to an authority. Nex never files reports for customers;
- (d) provide legal, regulatory or compliance advice;
- (e) use biometric data, facial recognition or emotion recognition, unless a feature is expressly described as doing so and the customer has enabled it under its own lawful basis; or
- (f) infer sensitive characteristics, such as race, religion, health or political opinions, from names or other data, except where a source explicitly states a political role needed for PEP screening.
3.2 Customer configuration. Customers must not configure or use the services so that AI output is applied automatically, without human review, to decisions about individuals.
4. Data Used by AI Features
4.1 Inputs. AI features process the data the customer submits for screening or analysis, the related screening results, and public-source content such as official lists, open-source PEP data and news articles.
4.2 Customer Data not used to train shared models. Nex does not use Customer Data, including names screened, case notes and documents, to train or improve AI models that are made available to other customers, unless the customer has given its prior written consent.
4.3 Service improvement. Nex may use aggregated and de-identified usage information, such as whether users accepted or rejected a suggestion, to measure and improve the quality of its features, in line with the Data Processing Agreement and the Privacy Notice.
4.4 Retention. Prompts and outputs generated in the services are stored as part of the customer's records for the period set out in the Data Processing Agreement, so that decisions can be audited.
4.5 Role of the parties. For personal data processed by AI features on the customer's behalf, the customer is the controller and Nex is the processor, under the Data Processing Agreement.
5. Accuracy, Limits and Customer Responsibilities
5.1 Not 100% accurate. AI output can be incomplete, outdated, wrong or misleading. It can miss relevant information, include irrelevant information, or state something with more confidence than is justified. Nex does not warrant that any AI output is accurate, complete or suitable for any particular purpose.
5.2 Customer responsibilities. The customer is responsible for:
- (a) making sure that trained staff review AI output before relying on it;
- (b) checking AI summaries and suggestions against the underlying sources;
- (c) deciding whether AI-assisted features are appropriate for its own risk assessment, regulatory obligations and model-risk governance;
- (d) documenting its decisions and reasons independently of the AI output; and
- (e) telling its own customers, where the law requires, that AI-assisted tools are used in its compliance checks.
5.3 Regulatory expectations. Regulators may expect firms to understand, validate and govern the tools they use. On request, Nex will provide reasonable information about how a feature works to support the customer's own validation, subject to confidentiality.
5.4 Liability. Nex's responsibility for AI features is subject to the warranties, disclaimers and limitations of liability in the Terms. A screening outcome or AI suggestion that the customer disagrees with is not, by itself, a defect in the services.
6. Third-Party AI Providers
6.1 Use of providers. Some AI features may use models or infrastructure provided by reliable, established third-party providers. Nex selects these providers through due diligence on security, privacy and reliability.
6.2 Contractual protections. Where a provider processes Customer Data, it acts as a Nex sub-processor under the Data Processing Agreement, and is bound by written terms that require confidentiality, security, and processing only on Nex's instructions. Nex chooses provider settings that do not allow the provider to use Customer Data to train its own models.
6.3 List on request. The current list of sub-processors, including AI providers, is available on request at privacy@nex.systems.
6.4 Provider changes. Nex may change providers to maintain quality or security. Changes to sub-processors are handled as set out in the Data Processing Agreement.
7. Governance, Testing and Oversight
7.1 Framework. Nex manages AI risk using principles drawn from the NIST AI Risk Management Framework and ISO/IEC 42001: govern, map, measure and manage.
7.2 Before release. Each AI feature is assessed for its intended use, foreseeable misuse, data protection impact and risk of harm, and tested for accuracy, consistency and bias on representative data, including names from different languages and regions.
7.3 After release. Nex monitors AI features for quality, drift, errors and user feedback, keeps logs needed to investigate problems, and reviews significant features at least once a year.
7.4 Incidents. If an AI feature produces materially wrong or harmful results, Nex will investigate, correct or disable the feature where needed, and inform affected customers where appropriate.
7.5 Human oversight by design. Features are designed so that users can see the basis for a suggestion, override it, and record their own decision and reasons.
7.6 Evolving law. Laws and guidance on AI are developing quickly. Nex monitors them and will update its features and this Statement as needed.
8. Questions, Feedback and Contact
8.1 Feedback. Users can flag an incorrect or unhelpful AI result inside the services. Nex reviews this feedback to improve its features.
8.2 Individuals. People whose names have been screened by a Nex customer should contact that customer, which controls the screening. Questions about Nex's own processing can be sent to privacy@nex.systems.
8.3 Changes. Nex may update this Statement by publishing a revised version at nex.systems/legal.
8.4 Contact.
| Purpose | Contact |
|---|---|
| AI questions, model information requests | privacy@nex.systems |
| Security concerns | security@nex.systems |
| Contractual notices | legal@nex.systems |
Nex Global LLC, trading as Nex Systems, Wyoming, United States.
Common questions
Short answers for orientation. The document above governs if anything differs.
In adverse-media filtering, name matching assistance, summaries, risk-scoring suggestions and a compliance assistant, presented in the product as Elodie AI; not every feature is on every plan. Sanctions list collection, list updates and core exact and fuzzy matching run on deterministic rules, not generative AI (see Section 1).