Terms of service

It is between Evidova, which this page calls Evidova, we or us, and the organisation whose workspace uses the service, which this page calls you. If you accept these terms on behalf of an organisation, you are confirming you may bind it.

Evidova is for business use. Workspaces are created by invitation, there is no public signup, and the service is not offered to consumers. Where an order form, pilot agreement or data processing agreement signed by both of us says something different, that document wins over this page.

You own the conversations you send us and the scores we produce from them. Nothing here transfers ownership of either.

You give us the narrow permission we need to do the job: to receive, store, process and display your data so the service can run, and to send conversation text to the model providers named in our privacy policy. That permission ends when your data is deleted.

We do not use your data to train, fine-tune or improve any machine-learning model. We may use de-identified aggregate statistics to improve our default checks and published benchmarks, and section 4 of the privacy policy sets out exactly what aggregate means and how to opt out.

Evidova reads conversations from the platforms you connect, scores them against checks you control, shows you where users gave up, and raises alerts when a measure crosses a limit you set. What you get depends on your plan and on which platforms you connect.

We improve and change the product. We will not remove a capability you rely on without telling you first.

Evidova produces opinions, not verdicts. A score is evidence for a human decision. It is not a certification, an audit opinion, an accreditation, or legal, regulatory or professional advice.

• The report you can export is a convenience export. It is dated, it names the data it drew on, and it does not certify that you comply with any law, standard or framework. Do not present it as one.
• Nothing we produce is a substitute for your own legal, compliance or clinical judgement.
• We publish how well our judges agree with human reviewers, because a number you cannot check is not evidence. Read those figures as a measure of quality, never as a guarantee about any single conversation.

The service is built on large language models. Their outputs may be inaccurate, incomplete, inconsistent between runs, or confidently wrong. Fabricated detail is a known characteristic of this technology, not a defect in ours.

You are responsible for reviewing a finding before you act on it. The product is designed for exactly that: every score opens onto the evidence beneath it, including where the judging panel disagreed with itself. Use it.

You must not use Evidova as the sole or decisive basis for a decision about a person.

That includes hiring, promotion, discipline, pay, rostering and dismissal, and it includes any decision about credit, insurance, housing, education, healthcare or a legal right. Where a score informs such a decision, a person must review it, that person must be able to reach a different conclusion, and you must meet whatever the law where you operate requires of automated decision tools, including notice, bias testing and record keeping.

Today the product scores bots, not people. If per-person quality becomes available, this clause applies to it from the first day and does not need to be rewritten.

• Having a lawful basis for the conversations you send us, and for the recording of them.
• Telling the people in those conversations that their conversation may be recorded and reviewed for quality, in whatever way the law where you operate requires.
• Deciding your own retention window, and switching redaction on before you connect a live platform if you need it. It cannot be applied to data that has already arrived.
• Keeping your API keys and your team’s access under control, and removing people who leave.
• The accuracy of what you send us. We score what arrives.

You hold the relationship with your end users, so these duties can only sit with you. We give you the tools; the basis is yours.

• Send us data you have no right to send, or content that is unlawful.
• Use the service to build a competing product, or to benchmark it for publication without our written agreement.
• Resell, sublicense or share your access outside your organisation.
• Attempt to break, overload or circumvent the service, or to reach another customer’s data.
• Remove or obscure the notices that say a figure is uncertain, a sample is small, or a panel disagreed.

The last one matters as much as the others. Those notices are the difference between evidence and a claim.

Fees, the billing period and the volumes you are entitled to are set by your order form. If you are on a free pilot, it is free, and it can end. We will give you notice before it does, and before any fee changes at renewal.

Invoices are payable within the period the order form states. Taxes are yours unless the order form says otherwise.

Each of us will protect the other’s confidential information, use it only for this agreement, and not pass it on except to people who need it and are under the same duty. This survives the end of the agreement. It does not cover information that is public, that the receiver already had, or that a law or court requires be disclosed.

We keep every right in the platform: the software, the models and prompts we build, our rubrics, benchmarks and documentation. You keep every right in your data and in your own checks.

If you send us feedback or an idea, we may use it to improve the product without owing you anything. That covers suggestions, not your data.

We promise to provide the service with reasonable skill and care, and to keep the security measures described in the privacy policy.

Beyond that the service is provided as is. We do not warrant that it will be uninterrupted or error free, that any score is accurate for a given conversation, or that it fits a purpose you have not told us about. Any uptime commitment must be written in an order form to exist.

Neither of us is liable to the other for indirect or consequential loss, lost profit, lost revenue, lost data beyond our duty to keep it, or loss of goodwill.

Our total liability under this agreement is capped at the fees you paid us in the twelve months before the claim arose.

Nothing in this section limits liability that the law does not permit to be limited, including liability for death or personal injury caused by negligence, for fraud, or for your duty to pay fees you owe.

You will cover us against claims that arise from the data you send us: that you had no right to send it, that it infringes someone’s rights, that the people in it were not told, or that a decision you took using a score broke a law. We will tell you promptly about any such claim and let you run the defence.

We may update this page. For a change that materially affects your rights or what we may do with your data, we will email the account owners at least 30 days before it takes effect, and if you do not accept it you may end the agreement and get a refund of fees already paid for the period after you leave.

We will not expand our rights over your data by editing this page quietly. If we ever need broader rights, we will ask.

• This agreement is governed by the laws of India, and disputes go to the competent courts in India.
• Neither of us is liable for a failure caused by something genuinely outside our control, though that does not excuse paying what is owed.
• You may not transfer this agreement without our consent. We may transfer it as part of a merger or sale of the business, and we will tell you.
• If a clause is unenforceable, the rest still stands.
• This page, plus your order form and any data processing agreement, is the whole agreement between us.
• Neither of us is the other’s agent, partner or employer.

Questions about these terms go to legal@evidova.com.