Responsible AI and Automated Systems Policy

This policy applies to every use of Verinex, whether via the Console, the API or any other interface. It supplements the Platform and Console Terms and the Acceptable Use Policy.

Verinex is designed so that advancement of capability is gated by governance, verification and evidence controls. The controls are architectural, not decorative: they are intended to make it possible to observe, inspect and defend how a result was produced. Users must not circumvent these controls or represent Verinex as a system that produces results without them.

Meaningful human oversight is a core assumption of Verinex. Deployers must ensure that an appropriately skilled person is able to understand, monitor and, where necessary, override or halt automated processes for their use case. Oversight must be proportionate to the risk of the use case and must be capable of being exercised in practice, not only in theory.

Verinex generates verification records and evidence artefacts that describe how a given result was produced. These records support inspection and audit. They do not substitute for the deployer's own assurance activities, professional judgement or regulatory obligations.

Automated systems, including Verinex, can be wrong. Outputs may be incorrect, incomplete, biased, out of date, unstable across runs or unsuitable for a particular context. Confidence signals, verification verdicts and evidence records reduce but do not eliminate uncertainty. Deployers must treat outputs as decision support, not as decision authority.

Verinex must not be used as the sole or determinative basis of a decision with significant legal or comparable effect on a person without meaningful human review. The high-impact use restrictions in the Acceptable Use Policy apply.

Verinex maintains logs, evidence records and verification artefacts as part of its ordinary operation. Deployers are responsible for retaining, controlling access to and, where relevant, disclosing such records in accordance with the law applicable to them.

The quality, representativeness and lawfulness of Customer Inputs materially affect Verinex's outputs. Deployers are responsible for the suitability of inputs for their intended purpose, including considering bias, coverage and provenance.

Prohibited and restricted uses are set out in the Acceptable Use Policy. They apply irrespective of any user interface or configuration choice that would appear to permit them.

Deployers must ensure that personnel operating Verinex have sufficient understanding of automated systems, their limitations and the specific characteristics of Verinex to exercise their responsibilities safely. This is a continuing obligation and should be reviewed as Verinex evolves.

Suspected incidents affecting the safety, integrity or reliability of Verinex — including unexpected outputs, apparent verification failures, evidence anomalies and suspected misuse — should be reported to Prodia Systems Limited without undue delay using the contact block below. Suspected security vulnerabilities should be reported under the Vulnerability Disclosure Policy.

Deployers using Verinex to interact with people, or to produce content or decisions affecting people, must provide transparency appropriate to their context, including the fact that automated systems are used and how outputs are reviewed.

Verinex does not warrant that outputs are accurate, complete, current, original, non-infringing, exclusive or fit for a particular purpose. Where downstream use is consequential, deployers must implement independent checks.

Automated processes can produce outputs that resemble third-party material or that reveal, by inference, information about training data or internal methods. Deployers must not use Verinex to reconstruct proprietary methods, confidential information or patent-sensitive mechanisms. See the Intellectual Property Notice and Patent Notice.

Adversarial inputs, prompt injection, model extraction attempts and manipulation of evidence flows are prohibited. Deployers must protect API credentials, isolate untrusted inputs and monitor for suspicious patterns of use.

The regulatory classification of a specific use of Verinex — for example under the EU AI Act — depends on the context in which the deployer places the system, not on Verinex in the abstract. Prodia Systems Limited does not represent that Verinex is categorically inside or outside any particular regulatory class. The deployer remains responsible for classifying its use case, completing any required conformity or impact assessment, and complying with the resulting obligations.

Nothing in this policy asserts external certification, accreditation or attestation of Verinex by any standards body or regulator. Any such certification, if and when obtained, will be published in the Security and Trust statement.