AI Services & Usage Policy

This policy explains how Bymond Private Limited approaches artificial intelligence: the AI services we build for customers, how data is handled when third-party AI providers are involved, and the limits of AI-generated output. It supplements our Terms & Conditions and Privacy Policy.

Where a signed agreement, statement of work or data processing agreement contains AI or data-handling provisions, that agreement governs the engagement and prevails over this page. Client contracts frequently contain stronger data-processing requirements than this general policy, and that is the right place for them.

Bymond provides AI development and AI automation services. In practice this usually means building systems around AI capabilities rather than training foundation models. For example:

• integrating AI models and APIs into applications and business workflows;
• building automation that classifies, extracts, summarises, drafts or routes information;
• designing retrieval and data pipelines that supply relevant context to a model;
• deploying and operating the infrastructure these systems run on, including self-hosted tooling;
• advising on where AI is a sensible fit for a problem, and where it is not.

Delivering AI capabilities often involves third-party AI infrastructure and model providers. Where a system sends data to such a provider for processing, that data leaves our infrastructure and is processed on the systems of that provider under its own terms.

We select providers for an engagement with regard to their terms, their data-handling commitments and their suitability for the workload, and we will tell you which providers a design depends on.

• Training. Where Bymond controls the applicable arrangement with an AI provider, we do not knowingly permit customer confidential information to be used by that provider for training its models, and we seek to use provider configurations and terms that support this.
• Purpose limitation. Customer data submitted for processing is used to deliver the agreed services and for no unrelated purpose. We do not use customer confidential data to build products for other customers.
• Where we are not in control. Where the customer holds the account with an AI provider, or directs us to use a particular provider or configuration, the terms of that arrangement apply and the customer is responsible for them. We will flag it if a requested configuration has data-handling consequences we think you should know about.
• Provider terms change. AI providers revise their terms, retention practices and model behaviour over time. We do not control those changes and cannot guarantee that a provider will keep its current terms indefinitely.

How personal data is handled more generally, including in the context of AI processing, is set out in our Privacy Policy.

• You must have the right to provide any data you submit, or instruct us to submit, for AI processing, including any consents, licences or permissions it depends on.
• You should not submit data into an AI workflow if doing so would breach a law, a contract with a third party, or a confidentiality obligation you owe to someone else.
• Where the data includes personal data, you remain responsible for the lawfulness of the purpose and the instructions you give us.
• You are responsible for reviewing AI-generated output before acting on it, and for deciding where human review is required in your own processes.

AI systems are probabilistic. They can produce output that is inaccurate, incomplete, outdated, internally inconsistent, or confidently wrong, and the same input can produce different output on different occasions.

• AI-generated output should be reviewed before it is relied upon, and particularly before it is published, sent to a third party, or used to make a decision.
• AI output is not legal, financial, medical, tax or other professional advice, and should not be presented as such.
• AI systems can reflect biases present in their training data. Where output affects people, that risk needs to be considered in the design of the process around it.
• Model behaviour changes when providers update or deprecate models. A system that behaved one way at delivery may behave differently later, which is one reason ongoing operation is worth planning for.

High-stakes use

Where AI output informs decisions with legal, financial, safety, employment or regulatory consequences, we recommend meaningful human review in the process rather than automatic action on model output. Where an engagement requires a particular level of oversight, accuracy or auditability, that should be agreed explicitly as part of the scope.

Our engineers may use AI-assisted development and productivity tooling in the course of their work. Where we do, the work remains subject to our normal review practices. A human engineer is responsible for what is delivered. We apply the same care to what is put into those tools as we do elsewhere: customer confidential information is handled according to the applicable agreement, and we take account of a tool's data-handling terms before adopting it.

AI features are subject to the same warranty and liability position as the rest of our services. For engagements governed by a negotiated agreement, that agreement controls. Otherwise the provisions in warranties and disclaimers and limitation of liability apply, including that we do not warrant that AI output will be error-free or suitable for every purpose, to the maximum extent permitted by applicable law.