odnoga is an AI routing, metering and governance layer. We do not develop, train or fine-tune general-purpose AI models, and we do not place models on the EU market under our own name. Models are provided by third parties (OpenAI, Anthropic, Google and others) listed in our sub-processor register.
1. Our role
- odnoga is not a provider of a general-purpose AI model within the meaning of Art. 3(63) AI Act.
- Where you use odnoga inside your own AI system, you are the provider or deployer of that system. The obligations attached to your use case (risk classification, human oversight, transparency towards affected persons) remain yours.
- odnoga acts as a supplier of an AI component and supports you with logging, traceability and documentation evidence.
2. What the platform gives you as evidence
- Automatic event logging (Art. 12/19 style record-keeping): every request records model, vendor, region, processing tier, token counts, latency, cost and outcome, retained for the period you configure.
- Model provenance: the catalogue records which vendor served each call, model version, deprecation status and replacement model.
- Change traceability: prompt versions, routing policies, budgets and admin actions are versioned and written to an audit log visible to your workspace admins.
- Human oversight hooks: budget caps, approval-gated budget changes, model allowlists and per-workspace concurrency limits.
- A per-workspace compliance summary in the product (Security & data) that reports your live region, retention mode, payload capture and data-subject request activity.
3. Transparency obligations
If your application generates or manipulates text, image, audio or video content, Art. 50 AI Act transparency duties (informing users they interact with an AI system, marking synthetic content) apply to you as the deployer. odnoga does not alter or strip vendor content markings and passes model identifiers back to you on every response.
4. Prohibited practices
Uses prohibited under Art. 5 AI Act — including social scoring, untargeted scraping of facial images, emotion inference in workplaces or education, and manipulative techniques causing significant harm — are forbidden on odnoga under our Acceptable Use Policy and will result in suspension.
5. High-risk deployments
If you deploy odnoga inside a high-risk AI system under Annex III, contact legal@odnoga.com. We can provide the technical information a supplier is expected to give under Art. 25(4): logging scope, retention configuration, region controls, sub-processor list and incident history.
6. Timing
Prohibited-practice rules apply since 2 February 2025; general-purpose AI model obligations since 2 August 2025; the bulk of high-risk obligations apply from 2 August 2026. We track these dates and update this statement as guidance and harmonised standards are published.
7. Contact
AI Act and compliance questions: legal@odnoga.com. Data protection questions: privacy@odnoga.com.