Responsible Data Work
Responsible practices for the data behind AI.
Training data and evaluation work can influence how AI systems behave. We approach that responsibility through clear project requirements, appropriate data sourcing, scoped human review, and transparent reporting about what we can support.
Data authorized for its use
Most work uses datasets and model outputs supplied or approved by clients. Where new material is collected, the source and intended use are agreed in advance. We review the available source and permission documentation within the project scope, record relevant origins when required, and flag unclear cases for the client to resolve. This review does not replace a legal determination about licensing, consent, or permitted use.
Privacy by default
We work with the information a task requires and no more. Where a project includes personal or confidential details, redaction or masking can be applied before or during the work. Access stays limited to the people assigned to your project.
Careful handling of sensitive content
Some evaluation and safety assignments may involve difficult or harmful material. We assess such requests before acceptance and take them on only where suitable reviewer capability, task specific guidelines, access controls, and appropriate handling arrangements can be confirmed. We may decline work if these conditions cannot be met.
Responsible contributor engagement
Where additional contributors are required, Omni Annotate AI evaluates their suitability for the assigned tasks and confirms the appropriate project arrangements. Contributors receive agreed instructions, task guidelines, and quality expectations. Access to client information is restricted according to the project's confidentiality and data handling requirements.
Honest data, honest reporting
We do not invent labels, scores, or results. Reports describe the review methods actually performed and their limitations. Where synthetic or AI generated material is part of the client's dataset, it is handled and labeled according to agreed criteria. Any proposed use of AI assisted tools in the delivery workflow is discussed with the client and must comply with the project's confidentiality, platform, and data use requirements.
Work we do not accept
We do not currently accept facial recognition or biometric identification annotation projects. We may also decline work whose purpose, data source, permissions, reviewer requirements, or technical needs cannot be confirmed or supported responsibly.
Raising a concern
If you have a concern about how data is handled on an active project, use the communication channel agreed for that engagement. For a new inquiry, use the Discuss Your Project page and avoid sharing confidential material until appropriate arrangements are in place.
Start With a Project Discussion
Tell us about your data, project objectives, and the annotation, training data preparation, or AI evaluation support you need. A short summary is enough. Please do not share sensitive or confidential data until appropriate confidentiality and secure data sharing arrangements are in place.
We’ll review your requirements, ask any necessary follow up questions, and propose a pilot or a calibrated first batch for smaller tasks tailored to your project.
The proposal will outline the task scope, sample volume, pricing, expected delivery timeline, and proposed acceptance criteria.
We review project inquiries and follow up to discuss requirements, scope, and appropriate next steps.