Skip to main content

Responsible AI Statement

Last Updated: 19 June 2026

Company: Myntriq Pte Ltd (UEN 202537571M), Singapore

Why We Wrote This

There are a lot of companies now publishing AI principles. Most of them say the same things: AI should be fair, transparent, accountable. These are the right words. They are also easy to write and difficult to honour.

We are writing this document because we think the responsibility question for AI has a specific answer in our context — and that answer is worth stating clearly, not abstractly.

Myntriq builds an AI operating system for small and medium-sized businesses across Southeast Asia. Our customers are not large enterprises with AI governance teams. They are the founders, operations managers, and department heads of companies that are trying to do more with less. When they deploy AI agents through MyntriqOS, the AI is acting on their behalf — in their name, in their business. That changes what accountability means, and it changes what we have to do to earn their trust.

The Accountability Question

When an AI agent in MyntriqOS drafts a customer email, qualifies a lead, or summarises the week's financials — who is responsible for what it says and does?

The answer is the customer. Always.

Not Myntriq. Not the AI model provider. The business owner who deployed the agent, configured it, and chose to act on its output.

This is not a liability disclaimer — it is the correct answer. The AI is operating in the customer's business, with the customer's data, on behalf of the customer's organisation. Responsibility follows control. And because responsibility follows control, control must be genuinely available.

This is why MyntriqOS is built the way it is:

  • Every agent operates at an autonomy level the customer sets
  • Every action is recorded in an audit log the customer can inspect
  • Every output can be reviewed before it is acted on
  • Agents can be paused, reconfigured, or overridden at any time

We are not trying to build AI that removes humans from the loop. We are trying to build AI that makes it worthwhile for humans to stay in the loop — because the AI is doing the work they would otherwise have had to do themselves.

What We Build For

Transparency over black boxes. If a MyntriqOS agent takes an action, the customer can see what it did, when, why, and at what cost. This is not optional — it is a core platform requirement. We will not ship agent capabilities that cannot be audited.

Human oversight as the default. When we configure agent autonomy levels, the default is always "ask before acting". Increasing autonomy is something the customer chooses, based on demonstrated performance, not something we push.

Model independence. We support multiple AI models — GPT-4o, Claude, Llama 3.3 — because no single model has a monopoly on being right. If a model changes, degrades, or is found to have systematic biases in a particular context, our customers are not locked in. They switch the model; they do not rebuild the business logic.

Data belonging to the customer. Customer content — conversations, documents, business data — belongs to the customer. We do not use it to train our own models. We do not sell it. We do not aggregate it to improve our product at the customer's expense. We process it to provide the service they have paid for.

What We Do Not Do

We do not allow AI agents to make autonomous decisions in contexts where human review is required — credit decisions, employment decisions, medical diagnoses, legal proceedings. These are areas where the consequences of error fall on individuals, not on the business. Our AI Usage Policy is explicit about this, and our platform configuration enforces it.

We do not deploy agents at autonomy levels that bypass the controls the customer has set. If a customer has configured an agent to ask for approval before sending emails, that configuration is enforced. We do not override it.

We do not generate content designed to deceive. We do not build tools for disinformation, synthetic identity, surveillance without consent, or any purpose whose value depends on the target not knowing it is happening.

We do not claim our AI is infallible. AI models make mistakes. They hallucinate facts, miss context, and reflect biases from their training data. Knowing this is not a reason to avoid using AI — it is a reason to use it with appropriate review, especially for consequential outputs. Our platform is designed on this assumption: not that the AI is always right, but that the human reviewing the AI's output can catch the cases where it is not.

How We Handle AI Errors

AI models make mistakes. MyntriqOS is designed with this as a given.

Agents are configured to flag uncertainty — rather than proceed confidently when they do not have enough information. When an agent is not confident, it should ask, not invent.

The Governance Dashboard gives administrators visibility over agent outputs before they are acted on, at the autonomy levels the customer has set.

Customers can roll back agent actions in supported workflows. Where rollback is not technically feasible, the audit log provides a clear record of what happened for the purposes of manual correction.

We maintain a channel for reporting agent errors and responsible AI concerns: hello@myntriq.io. Reports are reviewed and, where they reveal systemic issues with platform behaviour, are fed back into our engineering process.

How We Think About AI and Employment

This question comes up, and it deserves a direct answer.

MyntriqOS is designed for businesses that cannot afford to hire specialists. A 40-person distribution company does not have a dedicated marketing analyst, a sales development team, and a finance data analyst. MyntriqOS gives that company capabilities that would otherwise require those hires — not to replace existing employees, but to fill roles that were never filled.

For our typical customer, the relevant question is not "will this replace someone's job?" It is "will this give us the capabilities we need to compete?" We think the answer to that second question is yes. We also think that businesses that are better equipped and more organised create better conditions for the people who work in them.

We are aware that the broader AI employment picture is more complicated than this. We do not pretend otherwise. But we build for the customers we serve, in the context they operate in — and in that context, AI that works correctly is a tool for expanding what is possible, not reducing who is needed.

Our Commitments Going Forward

This document reflects where we are in June 2026. Our platform is live and being used by real customers. We are not a research lab — we are building a product, and product decisions involve trade-offs.

We commit to:

  • Publishing updates to this statement when our approach or platform capabilities change materially
  • Being honest about limitations, including our own certifications (which, as an early-stage company, are currently limited)
  • Maintaining the AI Usage Policy, Model Governance Policy, and Subprocessor List as live, current documents — not set-and-forget legal text
  • Responding to responsible AI concerns submitted to hello@myntriq.io

We do not commit to perfection. We commit to transparency about where we are, seriousness about the responsibilities that come with building AI agents that act in business contexts, and a genuine belief that AI can be a tool for expanding capability without surrendering control.

*Myntriq Pte Ltd — Singapore — June 2026*