1. Purpose and scope
We use artificial intelligence and automated methods to support parts of our maritime intelligence research, our platform operations and our internal work. This page explains how we control that use.
These controls cover our public content, our analyst work and our operational support. Where a client contract or an approved project procedure sets a stricter standard, we apply the stricter standard.
2. Where we use AI assistance
We use approved tools to organize source material, extract entities, translate and summarize text, draft content, detect patterns, classify records, correlate events and prioritize items for review.
Our platform also generates alerts from defined data conditions, and escalates an alert when the same condition persists across repeated checks. Escalation raises the priority of a review. It does not replace one. One of our analysts interprets every alert in its operational context before we act on it.
3. What AI does not decide
AI holds no command authority in our work. It does not make our final navigational, security, legal, sanctions, compliance, insurance, employment or commercial decisions, and it does not make them on behalf of a client.
We do not treat AI-generated text or a model score as a verified fact. Our responsible professionals review binding, safety-critical and client-facing work before we release it or act on it.
4. Sensitive data
We give each approved AI-assisted task only the data it needs. We remove or mask personal identifiers, client names, crew details and voyage identifiers when the task does not require them.
Client-confidential data, identity documents, crew lists, security plans, precise protected locations and proprietary voyage data are processed only in an approved environment, for a defined purpose, and under suitable contractual and security controls. We do not place that material in an unapproved public AI service.
5. Evidence and verification
We distinguish verified facts, official claims, third-party reporting, AIS observations and analytical possibilities. AI assistance never raises the reliability we assign to an unsupported source.
On vessel and incident work we check identity, date, position, source provenance and material inconsistencies. Where the evidence does not support a firm conclusion, we say so and state the level of uncertainty.
6. Human oversight
Every approved AI-assisted workflow has an owner, either a named team or a responsible professional. The owner defines the task, checks the output, manages escalation and can stop the workflow.
The people who approve a conclusion or an action remain accountable for it. Clients keep their own command, safety, legal and compliance responsibilities.
7. Error, bias and security
AI systems can omit context, invent detail, repeat the bias of a source or produce unstable results. We test higher-impact workflows, record their known limits and check the underlying sources before we rely on an output.
Access controls, confidentiality rules, incident reporting and supplier security reviews cover our AI-assisted work. We investigate material errors and data exposure, and we tighten controls when an issue reveals a weakness.
8. Suppliers
We assess each AI supplier in proportion to the data and the task involved. That assessment covers data use, retention, training settings, access control, hosting location, security, subcontractors and contract terms.
Supplier behavior and availability change over time. For work that cannot depend on a single external model or service, we keep human review and an alternative procedure.
9. Regulatory position
This page sets out how we work. It is not a legal classification. Each AI-assisted use may fall differently under the law depending on the system, its intended purpose, its users, the location and the level of risk.
We review this framework as our systems and the applicable rules change, including data protection law and the staged application of the European Union Artificial Intelligence Act. We use AI systems supplied by third parties. We are not a provider of a general-purpose AI model.
10. Questions and corrections
Write to info@marerisk.com, call +351 926 857 388 or +39 351 4607099 with a question about our AI use, a suspected error or a data concern. Include enough context to identify the page, output or workflow, without sending sensitive material by open email.