Skip to content
Relevant

Relevant · relevant.com.tr

Legal

Acceptable use policy

Relevant helps teams find and fix failures in their AI systems. This policy explains what you can and cannot do with the service so that it stays safe, reliable and lawful for everyone who depends on it.

In short: use it for evaluation, respect the data you handle, and do not attack the service.

Last updated:

Scope#

This policy applies to everyone who uses Relevant: workspace owners and members, people who call the API or SDK with a key, and anyone who connects an integration. It is part of the Terms of service. A customer is responsible for the activity of its members and keys.

The policy is written to protect the service, other customers and the people whose data passes through it. It is not meant to get in the way of normal evaluation work, including testing how your system behaves when it is pushed.

At a glance#

Encouraged
What Relevant is for.
  • Allowed: Evaluating and comparing LLM and retrieval systems you build or operate
  • Allowed: Importing production traces after removing data you do not need
  • Allowed: Safety and red-team testing of your own systems
  • Allowed: Automating runs from CI within your plan’s limits
Allowed with care
Fine when you meet the conditions in this policy.
  • Allowed with care: Datasets that contain personal data, under the data processing agreement
  • Allowed with care: Adversarial prompts and harmful model outputs used for testing
  • Allowed with care: Bring-your-own model provider keys, under that provider’s terms
  • Allowed with care: Automated traffic that approaches your plan’s limits
Not allowed
Grounds for restriction or suspension.
  • Not allowed: Uploading data you have no right to process
  • Not allowed: Accessing another workspace or bypassing access controls
  • Not allowed: Using the service to build or run unlawful or deceptive systems
  • Not allowed: Presenting a Relevant report as an independent audit or certification

Permitted use#

You may use Relevant to:

  • evaluate, diagnose and compare language-model applications and retrieval systems that you build, operate, or are authorized to assess;
  • import datasets and production traces, run evaluations with the judges you configure, and review ground truth with your team;
  • export results and reports, and share them inside your organization or with your own customers and auditors as your analysis;
  • call the API, SDK and CLI from your own systems and CI pipelines at the volumes your plan allows.

Data you upload#

You must have the right to upload and process everything you submit, and you remain responsible for it. When a dataset or trace contains personal data, Relevant acts as your processor under the data processing agreement.

Do not upload:

  • special categories of personal data, such as health, biometric, or political opinion data, unless you have agreed the arrangement with us in writing;
  • payment card numbers, government identifiers, passwords, private keys, access tokens or other credentials;
  • personal data of children, unless you have a lawful basis and the consent that applicable law requires;
  • content that is unlawful, infringes the rights of others, or that you are contractually forbidden to share.

Prohibited activity#

Abuse of the service

  • Probing, scanning or testing the vulnerability of the service without our permission. Report issues through the disclosure process instead.
  • Accessing, or trying to access, another workspace, another user’s data, or administrative functions you are not entitled to.
  • Circumventing rate limits, quotas, usage metering or role-based access controls.
  • Creating accounts or keys by automated means, sharing keys between unrelated organizations, or reselling access without an agreement.
  • Interfering with the service, for example by overloading it, introducing malware, or sending traffic designed to degrade it.

Unlawful or harmful use

  • Using the service in breach of applicable law, including data protection, export control and sanctions law.
  • Building, tuning or evaluating systems whose purpose is unlawful surveillance, deception, harassment, or discrimination against people.
  • Using Relevant’s judges or outputs to generate harmful content for distribution, as distinct from testing a system against it.

Misrepresentation

  • Presenting a Relevant evaluation, score or report as an independent audit, regulatory certification, or endorsement by Relevant.
  • Copying or scraping the service, its documentation or its outputs to build a competing evaluation product.

Red-teaming and adversarial datasets#

Testing a system with hostile input is part of responsible evaluation, so datasets may contain harmful, offensive or manipulative prompts, and model answers may contain content that would otherwise be restricted. This is allowed when:

  • the system under test is yours, or you are authorized to test it;
  • the content is used to measure and improve that system, not to create material for distribution;
  • you give reviewers fair warning and do not require anyone to review content they have not agreed to see.

Where you connect your own model provider account for judging or generation, that provider’s usage policy applies to those requests in addition to this one.

Fair use and rate limits#

Each plan includes a monthly volume of evaluated samples and a number of seats and datasets. Published limits are listed on the pricing page and API rate limits are in the documentation.

  • Free
    Samples per month
    2,000
    Seats
    1
  • Pro
    Samples per month
    25,000
    Seats
    3
  • Team
    Samples per month
    150,000
    Seats
    10
  • Enterprise
    Samples per month
    Custom
    Seats
    Unlimited

We may throttle traffic that threatens the stability of the service for other customers. Wherever possible we will contact the workspace owner first and suggest a plan, schedule or batching change that fits.

Reporting a violation#

If you believe someone is using Relevant in breach of this policy, write to security@relevant.com.tr with the workspace or content concerned and what you observed. We keep the reporter’s identity confidential where we can. For other questions, use the contact form.

Enforcement#

We act in proportion to the risk and give you the chance to put things right where we can:

  1. Review
    We look at the activity, the content involved and any context the workspace owner can give us.
  2. Notice
    We tell the workspace owner what the concern is and ask them to fix it within a stated period, which depends on the risk.
  3. Restriction
    If it is not fixed, we may limit the affected feature, throttle traffic, disable an API key, or remove specific content.
  4. Suspension or termination
    For serious or repeated breaches we may suspend or end access. Where the law allows, you will have 30 days to export your data.

We may suspend access immediately, without prior notice, when that is needed to stop imminent harm to people, to the service or to other customers, or to comply with the law. We will tell the workspace owner as soon as we can and explain why. Where required by law, we cooperate with competent authorities.

Questions and appeals#

If you think we restricted your workspace in error, reply to our notice or use the contact form. A person who was not involved in the first decision reviews each appeal, and we reply within five business days.

Changes to this policy#

We update this policy as the service and the law change. We announce material changes on this page and by email to workspace owners at least 30 days before they take effect. Continued use after that date means you accept the updated policy.

Contact#

This policy is issued by the entity below.

Legal entity
Relevant
Privacy and data protection
privacy@relevant.com.tr
Legal notices
legal@relevant.com.tr