AI transparency

This statement describes AI-assisted work used to develop and maintain the public 4SI website, research service, code and visual materials. It distinguishes internal assistance from the behavior of the live public service.

The public website does not call a generative-AI service when a visitor reads, searches or filters a page. Visitor browsing and filter activity are not sent to an AI model, and public conclusions are not generated on demand.

Approved AI-assisted tools may support source triage, field normalization, duplicate review, translation support, comparison, drafting, code generation, testing, accessibility review, design exploration and visual production.

Assistance is task-specific. It does not give a model authority to approve a source, establish a fact, assign Critical status, make a legal conclusion, publish an allegation or determine an outcome about a person or organization.

• AI output is not a source, fact or independent confirmation.
• Published factual claims must map to cited evidence.
• Missing facts are not filled through generation.
• Source facts, observed signals, validated patterns, 4SI assessments, modeled exposure and uncertainty remain distinct.
• Material calculations, links and source relationships are checked against the underlying record.
• Any publication of a significant allegation or assessment requires appropriate human review. Archived classifications retain their historical context.
• Corrections follow the published Corrections Policy .

Human review does not make model output infallible. It creates an accountable decision point and a correction path.

4SI may use generative tools for product visualization, architectural visualization, concept imagery, diagrams or controlled image modification. The website includes AI-generated and AI-modified product, packaging, lifestyle and architectural imagery. These scenes illustrate intended products and uses; they are not documentary evidence of a customer installation, a tested capability or an event involving the depicted people. Generated people do not provide endorsements. Where a visual could otherwise mislead viewers about a real person, place, event or deployment, it requires a clear disclosure with that visual; this general notice does not replace a required contextual disclosure.

Generated or materially modified visuals are not presented as incident evidence. Where technically supported and appropriate, 4SI preserves or adds machine-readable provenance or content credentials. Machine-readable metadata alone is not a sufficient human-readable disclosure and may be removed by delivery systems or downstream copying.

For AI-generated or manipulated text on a matter of public interest, 4SI applies human editorial review and responsibility and provides an additional disclosure where required by law or where omission could materially mislead the audience.

4SI maintains an internal record of approved AI providers, model or tool family, material task and review owner because providers and model versions can change. Public visitor analytics, contact-form contents, passwords, authentication secrets, unreleased vulnerability details and unnecessary personal data are not approved inputs to general-purpose AI tools.

Any use involving confidential, personal or security-sensitive information requires an authorized environment, a defined purpose, minimization, access control, retention review and any necessary contractual protection. AI providers do not receive public-site analytics merely because a visitor browses the website.

• no biometric identification of public-website visitors;
• no individual automated decision producing legal or similarly significant effects;
• no claim that AI output is objective, neutral, complete or error-free;
• no automated publication of unverified allegations;
• no live generative scoring during an ordinary page visit.

4SI reviews this statement when a material workflow, provider, public feature or legal requirement changes. The owner of the relevant work remains responsible for validation, publication and correction. Material incidents or failures are escalated according to their impact.

The EU AI Act’s Article 50 transparency obligations apply from 2 August 2026. 4SI assesses relevant transparency requirements against the official Regulation and applicable guidance for the use case. This statement does not represent that every form of AI assistance is subject to the same obligation. This statement describes current controls; it does not make a blanket claim of compliance with every AI law in every jurisdiction.

EU AI Act — official EUR-Lex text  (opens in a new tab) European Commission — Article 50 transparency guidelines  (opens in a new tab)