AI Disclosure

Noda reports are produced by an artificial intelligence system. Article 50 of the EU AI Act (Regulation (EU) 2024/1689) has applied since 2 August 2026. Article 50(5) requires that the information be given clearly, at the latest at the first interaction, and accessibly; this page and the in-product notices are how we do that. We would publish this page anyway, because a buyer deciding whether to spend money on a site is entitled to know how the answer was reached.

You define a site and choose what you want answered. A team of software agents then reads public and licensed sources — grid operator publications, planning and environmental registers, map and terrain data, meteorological datasets — runs deterministic engineering and financial calculations over what it finds, and writes a report.

The reasoning and the writing are done by a large language model. The engineering and financial computations are ordinary code, not model output: contingency screens, yield models, discounted cash flow. Where a number comes from a calculation, the report says so; where it comes from a source, the report names the source.

This is the most important sentence on this page. A report is generated end to end by the system and delivered to you without a Noda engineer reading it first. We do not present it as a reviewed professional opinion, and it is not one.

What we do instead is make the report checkable. Every figure carries an evidence class — user input, source-backed, operator-confirmed, modelled, proxy, or missing. Findings the system inferred are labelled as inferred and carry lower confidence than findings it read from a source. Where the data needed to answer a question does not exist publicly, the report says the data is missing rather than filling the gap.

A screening is not a connection offer and cannot become one. Before committing capital you must obtain, from the relevant grid operator or authority, the connection response, the queue position, and the reinforcement scope and cost. Yield figures based on modelled resource must be replaced by a measured campaign before they carry a financing decision.

The report names, for each open question, the specific document that would settle it. That list is the honest part of the product.

The system can be wrong on any individual finding. Confidence levels are calibration aids, not guarantees. Data coverage varies by country: a screening runs anywhere, but the depth of curated map layers and datasets differs between markets, and the report states where coverage was thin.

Where a source was ambiguous, the system reports the ambiguity rather than resolving it silently. Where two of its own calculations disagree, the report says which one is canonical and why.

If you find an output that is wrong, tell us at [email protected]. We would rather know.

Noda is the provider of this system within the meaning of Article 3(3). Customers who use it under their own authority are deployers within the meaning of Article 3(4).

The system is not a high-risk AI system. It is not a product, and not a safety component of a product, covered by the Union harmonisation legislation listed in Annex I, and it is not required to undergo third-party conformity assessment under that legislation (Article 6(1)). It does not fall within any limb of Annex III. In particular, it is not a safety component in the management and operation of critical digital infrastructure, road traffic, or the supply of water, gas, heating or electricity under Annex III(2): it produces an advisory document, it is not a component of any installation, and it neither controls, dispatches, protects nor operates any transmission or distribution equipment. It does not assess the creditworthiness of natural persons, determine access to essential services, or perform any of the other Annex III functions.

Article 6(3) is available only to systems listed in Annex III. Because this system is not listed, that derogation is inapplicable rather than declined, and the documentation and registration duties that attach to relying on it under Article 6(4) and Article 49(2) do not arise. The obligations in Chapter III for high-risk systems, and the general-purpose AI model obligations in Chapter V, do not apply to us. We do not claim conformity assessment, CE marking or registration that our classification does not require.

We record the reasoning on this page, we re-examine it when the product changes, and we will produce a fuller assessment to a competent authority on request. In Romania the market surveillance authority for Regulation (EU) 2024/1689 is ANCOM.

The intended purpose, and what happens if you change it. The intended purpose of this system, in the sense of Article 3(12), is to produce an advisory screening document about a prospective energy project for the customer who commissioned it. Use outside that purpose is not permitted under the Terms. Article 25(1) provides that a deployer who puts an AI system to a use that falls within Annex III becomes the provider of that use and takes on the obligations that go with it. If you were to use a report to assess the creditworthiness of a natural person, or as a safety component in the operation of a network, that would be your system, not ours.

Article 50(1) — interacting with an AI. Where you interact with an AI feature on our website or in the product, we tell you so at the point of interaction.

Article 50(2) — marking synthetic content. The narrative text in a report is machine-generated and is synthetic content within the meaning of Article 50(2). The figures, maps and tables are rendered deterministically from data and geometry by our own code, and are not AI-generated content.

Article 50(2) requires that synthetic content be marked in a machine-readable format. Under Regulation (EU) 2026/1744, which entered into force on 27 July 2026, systems placed on the EU market before 2 August 2026 meet that duty from 2 December 2026. This system falls in that group, and we will comply by that date by embedding provenance metadata in the exported file together with a visible statement on the report itself. The mechanism is described in the report's methodology annex from that date.

Until then the disclosure is made here, in the Terms and in the report's own methodology annex, and every report carries the identifier of the run that produced it, so any delivered document can be traced back to a specific generation.

Article 4 — AI literacy. The people who build and operate this system are trained on what it can and cannot do, and on the limits described on this page.

If you publish a report, a duty falls on you. Article 50(4) requires a deployer who publishes AI-generated text in order to inform the public on a matter of public interest to disclose that the text was artificially generated. The exemption is for text that has undergone human review with a person taking editorial responsibility for it — and as Section 3 says, nobody at Noda reviews your report. So if you file a report, or extracts from it, into a public consultation, a planning process or any other public process, that disclosure is yours to make — unless you review it yourself and someone in your organisation takes editorial responsibility for the publication, in which case the Article 50(4) exemption is available to you. As deployer you also carry the AI literacy duty in Article 4 for your own staff and contractors. Article 50(4) does not apply to us: reports are confidential documents delivered to the customer who commissioned them, not text published to inform the public on matters of public interest.

Report generation uses an external model routing provider and the model operator behind it. Those companies, what they receive, and the transfer safeguards are named on the Sub-processors page.

Two points that customers ask about. Your name, email address and company name are not sent with a report request; the request carries an opaque run identifier. And we instruct the routing layer not to retain or train on the content of our requests.

The system does not take decisions producing legal effects concerning any person, or similarly significantly affecting them, within the meaning of Article 22 GDPR. It assesses energy projects, not people.

The AI system changes as we improve it. Every report records the run identifier and the date it was generated, so a delivered report can always be tied to the behaviour of the system on that date. The date at the top of this page is the date of this disclosure, not of the system.

Questions: [email protected].