AI Transparency Notice
Pursuant to Article 50 of Regulation (EU) 2024/1689 (the EU AI Act)
Last updated: 2nd September 2026
Table of contents
Jump to section
1. Purpose of this notice
This AI Transparency Notice is published by Generation Impact Global SA (“we”, “us”, “our”) in accordance with Article 50 of Regulation (EU) 2024/1689 (the EU AI Act) and the applicable guidance issued by the European Commission in relation to the transparency obligations for certain AI systems under Article 50 of the AI Act.
The purpose of this Notice is to inform You, clearly and transparently, about: (a) which features of our Platform use artificial intelligence; (b) how those AI features work, at a general level; (c) what data is processed by AI features and how; (d) the limitations and risks of AI-generated output; and (e) Your rights and responsibilities when using AI features.
This Notice should be read together with our Terms and Conditions, Privacy Policy, and Data Processing Agreement, all available at www.generationimpact.global/legal.
2. Who we are
Generation Impact Global SA Rue de Lausanne 82, 1202 Geneva, Switzerland Registered under CHE-260.384.863 Email: [email protected]
Generation Impact Global SA is a company specialised in the delivery of a wide range of ESG solutions within a single digital platform, to help businesses achieve their corporate responsibility objectives.
We integrate third-party artificial-intelligence and document-processing services into our Platform. These currently include services supplied directly by OpenAI, LLC through the OpenAI API and document-processing services supplied by LlamaIndex Inc. through LlamaParse. Microsoft Azure OpenAI is used only for the Utility IQ invoice-extraction feature where that deployment is enabled. Depending on the relevant AI Feature and the circumstances specified in the EU AI Act, we may act as a deployer of a third-party AI system and may also have provider obligations, including where we place an AI system on the market or put it into service under our own name or trademark. We do not develop the underlying foundation AI models.
3. AI features in our platform
3.1 Overview
Our Platform incorporates a range of AI-assisted features (collectively, the “AI Features”). The AI Features are designed to support and augment professional decision-making in the fields of ESG data management, sustainability reporting, impact analytics, and resource and utility management.
The following table describes the functional categories of AI Feature currently available in the Platform. We do not list individual products or reporting frameworks because availability varies by Subscription and may change; the relevant feature is identified in the Platform when it is made available to You:
| AI Feature category | Description | Type of processing |
|---|---|---|
| Document parsing and extraction | Converts uploaded documents and files into parsed text or structured fields | Document parsing, OCR and information extraction |
| Disclosure assistance | Retrieves evidence and produces draft responses for individual sustainability disclosure requirements | Information retrieval and generative text |
| Utility invoice extraction | Extracts supplier, customer, account, consumption, cost, meter and billing-period information from utility invoices | Document extraction and structured-output generation |
The AI-assisted features available to You depend on Your Subscription plan or Order Form. We may add, modify, or remove AI Features from time to time. This Notice will be updated where a material change affects the categories of AI Feature, the recipients of Customer Content, or the way Customer Content is processed.
4. How the ai features work
4.1 General Logic
The AI Features use different processing paths depending on the function. Uploaded documents may first be sent to LlamaParse for parsing and OCR or processed using a local parser fallback. Relevant document text, data fields or evidence snippets may then be sent directly to the OpenAI API to extract information or generate draft disclosure responses. Utility IQ invoice extraction may use a separately configured Azure OpenAI deployment. The services analyse the submitted content and return parsed text, structured fields or draft responses. Customer Content is not used to train or improve the underlying general-purpose models unless GIG expressly opts in to such use.
4.2 What the AI Does
The AI Features perform the following types of processing: (a) document parsing and extraction: converting unstructured documents, PDFs, images, utility bills and invoices into parsed text or structured fields; (b) disclosure assistance: retrieving relevant evidence and producing draft responses for individual sustainability disclosure requirements; and (c) utility invoice processing: extracting supplier, customer, account, consumption, cost, meter, billing-period and related invoice information. The Platform also contains rule-based and other automated functions that are not treated as AI Features under this Notice.
4.3 What the AI Does NOT Do
The AI Features do not: (a) make autonomous decisions with legal or similarly significant effects on You or any individual; (b) assess, rate, or score companies, funds, or investments for regulatory purposes; (c) provide legal, financial, investment, or tax advice; (d) perform emotion recognition or biometric categorisation; (e) use subliminal techniques, exploit vulnerabilities, or engage in social scoring; or (f) use Your Content to train the underlying general-purpose models unless GIG expressly opts in. Certain features may write extracted or generated information directly into the relevant Platform workflow or allow multiple results to be accepted together. This does not remove Your responsibility to review the information before relying on it, publishing it or using it in an external disclosure.
5. AI risk classification
5.1 Our Assessment
Based on their intended purposes and the manner in which they are currently used, we have assessed the AI Features against the prohibited-practice and high-risk categories in the EU AI Act and the transparency obligations in Article 50. Our current assessment is that the AI Features are not used for a prohibited practice or a high-risk purpose listed in Annex III. Article 50 obligations may apply differently depending on the particular feature, output and use. We review this assessment when a feature, model, intended purpose or regulatory requirement changes. Specifically:
| Risk Category | Applicable? | Explanation |
|---|---|---|
| Prohibited AI (Article 5) | No | We do not deploy any prohibited AI practices, including social scoring, subliminal manipulation, exploitation of vulnerabilities, or real-time remote biometric identification |
| High-risk AI (Article 6, Annex III) | No | The AI Features do not fall within the high-risk categories listed in Annex III of the EU AI Act. They are not used for biometric identification, critical infrastructure, education, employment, law enforcement, migration, justice, or democratic processes |
| Limited-risk / transparency-risk (Article 50) | Feature- and use-dependent | Generative-text features may be subject to Article 50(2) where GIG has the relevant provider responsibility. A user-interaction notice under Article 50(1) may apply to a feature that directly interacts with a person where the AI nature of that interaction is not otherwise obvious. Document parsing and extraction are not subject to Article 50 merely because AI is used, but another transparency duty may apply depending on the output and intended use. GIG maintains and periodically reviews an internal feature-by-feature assessment. |
| Components that are not AI systems | No | Certain components (e.g. rule-based data validation) do not meet the definition of an AI system under Article 3(1) of the EU AI Act and are outside the scope of Article 50 |
5.2 Prohibited Practices Confirmation
We confirm that we do not deploy any AI systems that fall within the categories of prohibited AI practices under Article 5 of the EU AI Act, including but not limited to: (a) AI systems that deploy subliminal techniques beyond a person’s consciousness to materially distort behaviour; (b) AI systems that exploit vulnerabilities of specific groups; (c) social scoring; (d) real-time remote biometric identification in publicly accessible spaces; or (e) emotion recognition in the workplace or educational institutions.
6. Third-party processors and services
6.1 Per-Feature Provider Details
| Feature or processing step | Recipient | Processing location | Provider-side retention |
|---|---|---|---|
| Document parsing and OCR | LlamaIndex Inc. (LlamaParse) | United States for the currently configured endpoint; an EU endpoint is available only if separately configured | Source documents and parsing results may be cached for up to 48 hours after job completion unless caching is disabled; additional retention is governed by GIG’s agreement with LlamaIndex |
| Document extraction and draft disclosure responses | OpenAI, LLC through the OpenAI API | United States and other locations used by OpenAI and its subprocessors under the applicable OpenAI data terms and DPA | API abuse-monitoring logs may retain Customer Content for up to 30 days by default, unless approved retention controls apply or longer retention is legally required; Customer Content is not used for model training unless GIG expressly opts in |
| Utility IQ invoice extraction | Microsoft Corporation through a separately configured Azure OpenAI deployment | The geography configured for the applicable Azure deployment | Subject to the configured Azure service terms and abuse-monitoring controls; not used to train the underlying foundation model |
| SDG Mapper (external service outside the AI Features) | European Commission Joint Research Centre (JRC) | European Union / European Commission infrastructure | Subject to the terms and privacy information applicable to the JRC service; GIG does not state a fixed provider-side period where the service documentation does not specify one |
6.2 External SDG Classification Service
The SDG Mapper uses an external natural-language-processing classification service operated by the European Commission Joint Research Centre (JRC) to associate submitted document text with the UN Sustainable Development Goals. GIG does not treat the SDG Mapper as an AI Feature under this Notice. Submitted text is processed on European Commission infrastructure in the European Union under the terms and privacy information applicable to the JRC service. The JRC service is not an LLM provider, and its output is not described as generative AI output in this Notice.
6.3 Right to Change Provider
We reserve the right to modify, replace, or supplement the technologies used in the Services. We shall notify You of any material change to a sub-processor that affects the processing of Your Content in accordance with clause 5 of our Data Processing Agreement, including the applicable advance-notice and objection process. We shall also publish the updated Notice on our Website.
7. How your data is processed by AI features
7.1 Data Flow
When You use an AI Feature, the processing path depends on the feature: (a) for document parsing and OCR, the uploaded document may be transmitted over an encrypted connection to LlamaIndex Inc.’s LlamaParse service, or handled by a local parser fallback; (b) for extraction and draft disclosure responses, relevant document text, structured fields or evidence snippets are transmitted over an encrypted connection directly to OpenAI, LLC through the OpenAI API; and (c) Utility IQ invoice extraction may transmit invoice content to a separately configured Azure OpenAI deployment. The applicable service returns parsed text, structured fields or generated responses to the Platform. The Platform may store parsed text, document chunks, embeddings, summaries, evidence snippets, extracted or generated responses and provenance records as product data so the enabled feature can operate and its results can be traced. Provider-side retention is described in Section 6.1. Customer Content is not used to train or improve the underlying general-purpose models unless GIG expressly opts in to such use.
7.2 Data Minimisation
We apply data minimisation principles to AI processing, but the amount transmitted varies by processing step. A complete uploaded document may be sent to LlamaParse where full-document parsing or OCR is required. A complete document or extracted document text may also be processed where the requested extraction operation requires it. Later retrieval and synthesis steps are designed to use only relevant requirements, fields or evidence snippets. We do not represent that Customer Content is anonymised or pseudonymised before every transmission. Customers should therefore avoid uploading Personal Data that is not necessary for the relevant feature.
7.3 Personal Data
You should avoid including unnecessary Personal Data in Content submitted to the AI Features. Documents and invoices may nevertheless contain Personal Data. Utility IQ invoice extraction may process customer or account-holder names, addresses, tax identifiers, account or customer identifiers and telephone numbers where present in the source invoice and requested by the extraction schema. Such Personal Data is processed in accordance with our Privacy Policy and Data Processing Agreement and is not used to train the underlying general-purpose models unless GIG expressly opts in.
7.4 Logging and Audit
We log use of the AI Features for operational purposes, including usage tracking, error monitoring, security and performance analysis. Logs normally include timestamp, user ID, feature and operation used, usage consumed, model or provider information and error codes. Depending on the feature and error condition, logs may also contain document file names, provider job identifiers, provider error messages and, in limited parse-failure cases, returned model content. Access to logs is restricted to authorised personnel. Logs are retained for twelve (12) months.
8. Labelling and disclosure of AI-generated content
8.1 In-Platform Labelling
We provide transparency information and interface notices appropriate to the relevant AI Feature and its use, including where required by Article 50 of the EU AI Act. The measures available may include:
(a) an AI indicator, notice or other contextual information informing You that a feature is AI-assisted or that content has been generated or materially assisted by AI, where required by applicable law and appropriate to the relevant feature;
(b) for generative-text features, an on-screen indication or accompanying notice that the output is AI-generated or AI-assisted and must be reviewed before use; and
(c) for document-parsing and extraction features, review prompts and, where supported by the relevant feature, source references or verification indicators. The specific presentation may vary by feature and interface.
8.2 Machine-Readable Marking
Where GIG is responsible for an AI system within the scope of Article 50(2) of the EU AI Act, we take measures to ensure that covered outputs are marked in a machine-readable format and are detectable as artificially generated or manipulated, to the extent technically feasible and subject to the characteristics of the relevant output and generally acknowledged technical standards. The available marking method may vary by output format and may include metadata, provenance information or another supported technical mechanism. This paragraph does not imply that every output format supports persistent metadata or that Article 50(2) applies to every AI Feature.
8.3 Your Obligation to Disclose
When You use AI-generated Output from the Platform in external communications, regulatory filings, or publications, You are responsible for disclosing that such content was generated or materially assisted by artificial intelligence, where required by applicable law. This obligation is set out in Section 4.8 of our Terms and Conditions.
9. Limitations, risks, and accuracy
9.1 AI Output Is Not Guaranteed to Be Accurate
AI-generated output may contain errors, inaccuracies, omissions, or “hallucinations” (plausible-sounding but incorrect information). The quality of AI output depends on the quality, completeness, and clarity of the Content You provide. We make no representations or warranties regarding the accuracy, completeness, or fitness for purpose of any AI-generated output.
9.2 Known Limitations
The AI Features have the following known limitations:
Document parsing and extraction: (a) Results depend on the quality, format, layout and language of the source document. Poorly scanned, handwritten, non-standard, multi-page or heavily formatted documents may produce lower-quality results. (b) Extracted values, including names, identifiers, consumption data, cost figures, meter readings, currency, units of measure and reporting or billing periods, should be verified against the source document.
Disclosure assistance: Retrieved evidence may be incomplete or irrelevant, and generated draft responses may misstate or omit information from the source material or applicable disclosure requirement.
Generated responses: Generated text may not fully reflect Your source documents, regulatory obligations, jurisdiction or industry context.
All AI Features: (a) The AI model’s training data has a knowledge cutoff and may not reflect the very latest regulatory developments. (b) The AI model may produce different outputs for the same input on different occasions (non-deterministic behaviour). (c) AI-generated output is not a substitute for professional judgement.
9.3 Human Oversight Required
Human oversight varies by feature. Some workflows present individual results for review; some permit multiple results to be accepted together; and some may save extracted or generated information directly into a Platform workflow. Regardless of the workflow, You must review relevant source evidence and verify AI-assisted information before relying on it, publishing it, submitting it in a regulatory filing or using it to make a professional decision. The AI Features are designed to assist and augment, not replace, human judgement. You are solely responsible for decisions and external communications based on AI-assisted information.
9.4 Not a Substitute for Professional Advice
AI-generated output from the Platform does not constitute, and must not be treated as, legal advice, financial advice, investment advice, accounting advice, tax advice, or compliance certification. You are solely responsible for seeking independent professional advice where appropriate.
10. AI usage and limits
10.1 Usage-based access
Access to the AI Features is subject to usage limits determined by Your Subscription plan or Order Form. Usage of the AI Features is metered by the Platform, and Your current consumption and remaining allowance are displayed to You within Your Platform dashboard, so that You can monitor Your usage at any time.
10.2 Plan allowances
Freemium subscriptions include a limited AI usage allowance intended for evaluation and light use. Starter subscriptions include a broader allowance. Subscriptions agreed under an Order Form include the allowances set out in that Order Form. The allowances applicable to each plan are published on our pricing page at www.generationimpact.global/pricing and are further described in our Terms and Conditions.
10.3 Changes to allowances
The AI Features are subject to ongoing development. We may adjust the usage allowances applicable to each plan from time to time to reflect changes in the AI Features, the underlying AI technologies, or their operating costs. Where a change would reduce the allowance applicable to Your current Subscription, we shall notify You in accordance with our Terms and Conditions.
10.4 Reaching a usage limit
Where You reach the usage limit applicable to Your plan, the AI Features will be temporarily unavailable to You. You may resume use of the AI Features in any of the following ways:
(a) by waiting until Your allowance resets at the start of the next monthly usage period, at no additional cost;
(b) where such an option is made available, by purchasing additional AI usage through the Platform. Additional AI usage is chargeable at the prices published on our pricing page and is subject to the terms set out in our Terms and Conditions; or
(c) by upgrading Your Subscription to a plan with a higher usage allowance.
Your current usage, Your remaining allowance, and the date on which Your allowance next resets are displayed in Your Platform dashboard. Reaching a usage limit does not affect Your access to non-AI functionality of the Platform, or to Content and outputs already processed.
10.5 Usage records
We log usage of the AI Features as described in Section 7.4. Usage records relate to the volume and type of use and do not include the Content You submit to the AI Features or the output returned.
11. Your rights
In relation to the AI Features, and where provided by applicable law, You have or may have the right to:
(a) be informed when You are interacting with an AI system or exposed to AI-generated content, where required by applicable law. This Notice provides general information, and we provide additional in-Platform information or labelling for the particular AI Features and outputs for which such notice is required;
(b) access, rectify, or delete Your Personal Data processed by the AI Features, in accordance with our Privacy Policy;
(c) object to automated decision-making that produces legal or similarly significant effects, where such right applies under applicable data protection law (Article 22 EU/UK GDPR; Article 21 Swiss FADP) (note: the AI Features do not make such decisions; see Section 4.3);
(d) request information about the general logic, purpose, and limitations of the AI Features (this Notice provides such information; further details are available upon request);
(e) report concerns or complaints about the AI Features to us at [email protected]; and
(f) lodge a complaint with the competent supervisory authority in Your jurisdiction where applicable.
12. AI literacy
In accordance with Article 4 of the EU AI Act and to the extent applicable to us, we take measures to ensure a sufficient level of AI literacy among our staff involved in the development, deployment, and operation of the AI Features. These measures include regular training on AI capabilities and limitations, data protection in the context of AI, and responsible AI principles.
We encourage You to ensure that Your Authorised Users who interact with the AI Features are appropriately informed about: (a) what the AI Features do and do not do; (b) the limitations described in Section 9; (c) the need to verify AI-assisted information before reliance or external use, including where a workflow permits bulk acceptance or automatic application; and (d) their obligations regarding disclosure of AI-generated content.
We make available onboarding materials, product documentation, and, for customers subscribing under an Order Form, training sessions to support AI literacy among Your users.
13. Responsible AI principles
Our use of AI within the Platform is guided by the following principles, aligned with the EU AI Act and the Council of Europe Framework Convention on AI:
Transparency: We clearly disclose when AI is used, how it works, and what its limitations are.
Fairness and non-discrimination: We design AI features to avoid bias and discriminatory outcomes. We monitor for and address potential biases in AI output.
Privacy and data protection: We process data in accordance with applicable data protection laws. Customer Content is not used to train the underlying general-purpose AI models unless GIG expressly opts in to such use.
Accuracy and reliability: We implement quality controls and encourage human oversight of all AI-generated output.
Human oversight: We support human review appropriate to the relevant feature and require customers to verify AI-assisted information before external reliance, publication or submission.
Security: AI processing uses encrypted connections, access controls, and secure infrastructure.
Accountability: We maintain logs, audit trails, and documentation to support accountability for AI-assisted processing.
14. Applicable regulatory framework
This Notice has been prepared with reference to the following regulatory instruments, to the extent applicable:
| Regulation | Applicable Provisions |
|---|---|
| EU AI Act (Regulation (EU) 2024/1689) | Article 4 (AI literacy); Article 5 (prohibited practices); Article 50 (transparency obligations for providers and deployers) |
| Council of Europe Framework Convention on AI | Transparency, human rights, data protection, and non-discrimination in AI |
| Swiss FADP | Data protection obligations applicable to AI processing of Personal Data; Article 21 (automated individual decisions); Article 24 (notification of data security breaches) |
| EU GDPR | Articles 13, 14, 22: transparency, automated decision-making rights |
15. Updates to this notice
We may update this AI Transparency Notice from time to time to reflect changes in our AI features, regulatory requirements, or operational practices. We shall publish the updated Notice on our Website and, where the changes are material, notify You by email. The “Last updated” date at the top of this Notice indicates when the most recent revision was made.
16. Contact us
If You have any questions about our use of AI, wish to request further information about the AI Features, or wish to report a concern, please contact us at:
Generation Impact Global SA
Rue de Lausanne 82, 1202 Geneva, Switzerland
All enquiries, including AI, data protection and platform support: