AI transparency

In plain English

What it does. Metis is an AI assistant. It retrieves, quotes and writes answers — from a defined regulatory corpus and from the records you put in yourself. Answers using retrieved sources provide references for you to open and inspect.

What it does not do. It does not make the final professional or safety decision. Scheduled research can run unattended within a restricted tool set; actions requiring interactive approval are refused during those runs. Your work is not used to train models.

Who is responsible. You are. Every answer is informational only — verify it against the approved exposition or the competent authority before you act on it.

What it holds, and removing it. You can inspect and edit your saved memory, export Brain notes, and request account deletion. Shared records and retention exceptions are described in the Privacy Policy.

Regulation (EU) 2024/1689—the EU AI Act—requires clear disclosure for AI systems that interact with people and sets marking or labelling duties for certain generated content. The Article 50 transparency rules have applied since 2 August 2026, and the European Commission adopted its final Article 50 guidelines on 20 July 2026. Avioverse treats those requirements as a design requirement.

Where you meet AI in Avioverse

The assistant

The chat in the app, browser extension side panel and optional messaging channels are AI assistant surfaces. At or before the first interaction, each surface tells you that you are interacting with AI. The assistant must not present itself as a person.

Document intelligence you invoke

OCR, certificate autofill, records decomposition, document ingestion and structured extraction can use AI to read material you provide and prepare fields for review. The feature should identify a draft or proposed result before it becomes a relied-on record.

Drafting aids you invoke

Procedure drafts, audit summaries, proposed actions, risk-assessment drafts and other writing aids generate working material. You review, edit and decide whether to use it.

Document summaries

Short plain-language summaries of Airworthiness Directives, Safety Information Bulletins and some library documents are written by AI from the document text. Each summary is marked where you read it, as described below. The directive, bulletin or publication itself remains the source to rely on.

The disclosure you should see

The chat in the app and the browser extension shows this notice under the message box, before your first question. The optional Telegram channel sends it once, in the message that confirms the connection:

You're chatting with an AI assistant. Answers can contain errors — verify against official sources before acting on them.

The Metis mark (), the letter M drawn on a network of nodes, is Avioverse's symbol for AI. It marks the assistant and every AI feature in the app. AI-written summaries shown outside a chat carry the mark, and the note under each summary starts with the same mark and reads:

Summarised by AI.

Where AI-generated content has no single source document, such as insights Metis adds to a Brain note, the disclosure is:

AI-generated. Informational only — not a substitute for the approved exposition or the competent authority.

Accuracy, grounding and human review

AI can invent facts, misunderstand context, omit a condition or rely on an outdated or inapplicable source. Avioverse reduces this risk through source-grounded retrieval, citations, structured validation and review gates, but those measures do not make output infallible.

You remain responsible for checking current official sources, approved data, local procedures and the competent authority before acting. The person with the required competence and authority must make the final professional or safety decision.

What AI does not decide

Avioverse AI does not:

issue an approval, certificate, licence, release to service or legal determination;
determine aircraft airworthiness or whether an AD is definitively applicable;
replace an approved exposition, programme, record system or safety-management process;
make automated decisions about a user with legal or similarly significant effects; or
use emotion recognition or biometric categorisation.

Generated-content marking

Article 50(2) requires providers of in-scope generative AI systems to make generated output machine-readable and detectable as artificially generated or manipulated, as far as technically feasible. The European Commission and AI Board confirmed in July 2026 that the voluntary Code of Practice on Transparency of AI-Generated Content is an adequate framework for demonstrating the Article 50(2), (4) and (5) obligations.

Avioverse records generation provenance inside the Service and uses the visible disclosures described above where you meet AI. Avioverse builds on AI models from specialist vendors and relies on their marking: text carries a vendor's invisible watermark where that vendor applies one. Files Avioverse produces carry no AI label or metadata, though text a model wrote for them may carry that vendor's watermark. Text and files you take out of the Service — a copied answer or a downloaded document — should not be assumed to carry an AI marker; the record inside the Service remains the provenance source.

Avioverse is not currently a signatory to the Code of Practice. We publish a gap analysis against each of its required measures. Signing remains open and may be reconsidered. Signing is voluntary; compliance with Article 50 is not. The operator records the feature-by-feature role, marking and accessibility check and repeats it when features change.

Your publication responsibilities

If you publish AI-generated or manipulated text to inform the public on a matter of public interest, Article 50(4) may require disclosure that it was AI-generated. The AI Act provides an exception where the text has undergone human review or editorial control and a natural or legal person holds editorial responsibility. Other professional, contractual or sector rules may still require disclosure or approval. This is general information, not legal advice for your organisation.

Our published material

Avioverse articles and implementation guides identify a human author or reviewer who holds editorial responsibility. AI assistance does not remove that responsibility. Editorial responsibility rests with Dionysis Kefalas, Founder of Avioverse, reachable at the contact address at the end of this page. Verbatim regulatory quotations remain separate from Avioverse explanation and are tied to their source.

Your data and model providers

Avioverse may select different third-party model providers by task. Relevant prompts, documents and context are processed to provide the feature you request. Avioverse does not use your content to train AI models and requires service providers to process it for providing the contracted service, subject to the safeguards in our Privacy Policy.

Legal references and contact

Questions or reports about an AI disclosure can be sent to privacy@avioverse.io.