Start Excel trial

EU Regulation 2024/1689 disclosure · yigcore.com

AI transparency

Effective

This page documents our self-classification under EU Regulation 2024/1689 (the "AI Act"). It is published in good faith based on the product as currently scoped. If you intend to use YiG for any of the high-risk use cases listed in Terms section 6, contact us first.

YiG is an AI system within the meaning of EU Regulation 2024/1689 (the "AI Act"). This page sets out how we classify YiG under the AI Act, the obligations we accept as provider, the use cases we forbid, and how AI-generated output is marked.

1. Identity of the provider

The provider of YiG within the meaning of AI Act Art. 3(3) is the founder of YiG, acting as a natural person pending formal incorporation. Contact details are in our Legal Notice. Compliance contact: [email protected].

2. Self-classification — limited-risk AI system

We classify YiG as a limited-risk AI system under the AI Act. The system's intended purpose is to draft reconciliations, variance bridges, and audit-ready evidence artefacts for finance teams. The intended use does not fall under any high-risk category listed in AI Act Annex III, and does not involve the prohibited practices of Title II.

This classification depends on our customers' actual use respecting the forbidden use cases listed in section 5. If a customer wishes to deploy YiG for a high-risk use, the system moves into the high-risk regime of Title III, which requires a conformity assessment, technical documentation, post-market monitoring, and CE marking. We do not currently support high-risk deployments and ask customers with such intent to contact us before joining the waitlist.

3. GPAI status

YiG is not a general-purpose AI model ("GPAI") within the meaning of AI Act Art. 3(63). YiG is an AI-based application that calls a third-party inference endpoint chosen by the customer (the "bring-your-own-key" model). The customer contracts directly with the model provider (Anthropic, OpenAI, Microsoft Azure, or a self-hosted endpoint) and the GPAI obligations of AI Act Title V attach to the upstream model provider, not to YiG.

4. Transparency obligations under Art. 50

Article 50 of the AI Act sets two transparency obligations directly relevant to YiG:

  • Art. 50(1) — interaction notice. Natural persons interacting with YiG are informed they are interacting with an AI system. This is provided through the product UI ("AI drafts your close. Your team ships it." on the homepage; "YiG draft." labelling on each bridge-item narrative in product surfaces).
  • Art. 50(2) — AI-generated content marking. YiG outputs that are AI-generated text, structured data, or analysis are marked as such in the product surfaces. Reconciliation drafts carry a per-item YiG-generated marker in the audit log (event=tool.execute.complete tool=finance_basis_reconcile) and in the reviewer surfaces (Slack Canvas card header; Office add-in review pane header).

5. Forbidden use cases

Use of YiG — or incorporation of YiG outputs into any downstream system you operate — is forbidden for the following purposes. These categories correspond to AI Act Annex III (high-risk uses) and Title II (prohibited practices). They are also reproduced in our Terms.

  • Creditworthiness evaluation or credit scoring of natural persons;
  • Risk assessment or pricing of life or health insurance for natural persons;
  • Recruitment, candidate screening, performance management of natural persons in employment;
  • Educational or vocational scoring, admission, or evaluation of natural persons;
  • Assessment of eligibility for public assistance, social benefits, or essential public services;
  • Safety components of critical infrastructure (energy, water, transport);
  • Law enforcement, border control, migration, or asylum decision-making;
  • Subliminal, manipulative, or exploitative techniques targeting natural persons (AI Act Art. 5);
  • Real-time biometric identification, social scoring, or emotion recognition (AI Act Art. 5).

6. Human oversight

YiG is architected so that no output produced by the AI component reaches a customer-controlled artefact (e.g., excel_write.complete in the audit log) without a preceding human-reviewer approval event (draft.transition with to=approved). This is enforced at the architecture layer, not at the application layer, and is documented in /docs/security-and-data-handling.

7. Training data & intellectual property

YiG is not a model and is not trained on customer data. The upstream model (Anthropic, OpenAI, Azure, or self-hosted) whose endpoint YiG calls is selected by the customer under their own contract. We do not retain customer prompts or outputs in any YiG-controlled training corpus.

8. Incident reporting

Serious incidents involving YiG's intended use (within the meaning of AI Act Art. 73) can be reported to [email protected]. We will acknowledge within 5 business days and report to the relevant supervisory authority within the timeframe required by the AI Act once it is fully in force and a competent authority is designated in the founder's jurisdiction.

9. Changes

This page will be updated as the AI Act regulatory framework develops, as our intended use evolves, or as we onboard customers in adjacent verticals. Material changes will be announced by email to waitlist members at least 15 days before taking effect.

10. Contact

AI Act questions, classification queries, high-risk use discussions: [email protected].