This page covers work involving artificial intelligence: assistants and chatbots, document and data extraction, content drafting, and automation built on AI models. It sits alongside our Design & Service Policy rather than replacing it.
What AI can and cannot be relied on to do
AI systems produce likely output, not verified output. They can be fluent, well formatted and wrong in a single detail. We design around that rather than pretending otherwise, but no configuration removes it entirely.
We do not represent that any AI feature will be free of error, and we recommend AI only where the output can be checked, either automatically against a rule or by a person before it is acted on.
Human review
Unless expressly agreed in writing, AI output is treated as a draft or a suggestion. You remain responsible for reviewing output before it is published, sent to a customer, relied on for a decision, or used in anything with legal, financial, medical or safety consequences.
Where we build an approval step into a system, that step is there to be used. Removing or bypassing it is your decision and your responsibility.
Your data and your material
Assistants and extraction pipelines are grounded in material you provide. You confirm that you hold the rights to that material and that supplying it to us, and to the model provider used, does not breach any obligation you owe to anyone else.
We will tell you which provider a system uses and what data leaves your systems. If you have a requirement that data must not leave a particular jurisdiction or provider, tell us before the work is scoped, because it affects what can be built and what it costs.
Third-party models and services
AI features usually depend on a third-party model provider. Those providers change their models, pricing, rate limits, terms and availability on their own schedule and without our involvement.
We are not responsible for a provider changing or withdrawing a model, altering its output, changing its pricing, or suspending service. Where such a change requires work to adapt your system, that work is chargeable. Where a provider's own terms apply to your use, those terms apply in addition to ours.
Running costs
AI features usually carry a usage-based cost billed by the provider. We will estimate it before the work begins, based on expected volume. Actual cost follows actual usage, which we do not control. Where we hold a provider account on your behalf, those charges are passed on at cost and are payable in the same way as any other third-party cost.
Accuracy, bias and content
AI output can reflect bias present in training data and can produce content that is inaccurate, outdated or inappropriate for your audience. Where output is customer facing, we recommend and can build guardrails, but responsibility for what your business publishes remains with your business.
You are responsible for ensuring that AI-assisted content you publish complies with advertising, consumer protection, professional and industry rules applying to you.
Confidentiality
Do not put information into an AI feature that you would not be willing to share with the model provider. Where confidentiality requirements are strict, tell us, and we will scope accordingly, including options that keep processing within systems you control.
Liability
To the fullest extent permitted by Malaysian law, we are not liable for loss arising from reliance on AI output, from decisions taken on the basis of AI output, or from a third-party provider's acts, changes or omissions. Our overall liability limit in the Design & Service Policy applies to this work as it does to everything else.
Our commitment
We will tell you honestly where AI is a poor fit, including where the honest answer is that a simpler tool, or no tool at all, will serve you better. We would rather decline a project than sell one that will not repay its cost.
The signed quotation or agreement for your project takes precedence wherever it differs from this page.