Responsible AI Policy

Every system we build or recommend includes a defined role for human judgment proportionate to the stakes: approval gates for consequential actions, escalation paths for uncertainty, and clear accountability. We decline engagements where a client requires fully opaque autonomous decisions about people.

Client data shared with us is used only for that client's deliverables. It is never used to train or fine-tune models — ours or anyone else's — and we contractually flow this restriction down to every model provider we engage.

We take no referral fees, reseller margins, or partner incentives from AI providers or platforms. Recommendations are made on benchmark evidence and exit-cost analysis, documented in writing for every engagement. See our principles.

Nothing ships on vibes. Systems we deliver include automated evaluation harnesses with pre-registered baselines, adversarial (red-team) testing before launch, and regression gates on every model change.

For systems that affect people — hiring, credit, support, healthcare adjacency — we test for disparate performance across relevant groups before launch and monitor for drift after it. Where disparities appear, we surface them honestly rather than tune them quietly.

• Clients always know which models processed their data, in what jurisdiction.
• AI-generated content we publish is human-reviewed; material client deliverables document AI involvement per contract.
• We do not build systems designed to deceive people into believing they are talking with a human when the client's use case requires disclosure.

Our governance work maps to the EU AI Act, GDPR, India's DPDP Act 2023, and applicable sector regimes (financial services, healthcare, energy). We track obligations across jurisdictions our clients operate in and flag changes proactively under quarterly reviews.

We will not accept engagements for: surveillance of individuals without lawful basis, manipulation or dark-pattern applications, autonomous weapons, unlawful profiling, or systems whose primary purpose is generating disinformation.

Internally, we follow the same rules we set for clients: approved tool lists, no confidential client material in unapproved tools, and annual responsible-AI training for every consultant and engineer.

This policy is owned by Moomoth Media Private Limited's leadership and reviewed at least annually. If you believe work bearing our name has violated it, write to [email protected] — reports are reviewed by someone outside the originating engagement, and we respond within ten business days.