Responsible AI Use Policy
How UIT Software governs the AI it builds and the AI it uses internally — where human oversight applies, what our systems are permitted to do on their own, and what they never do.
Purpose
UIT Software builds AI-native software for healthcare organizations — systems that touch clinical workflows, patient data, and decisions with real consequences. This policy exists because that context demands an explicit, public statement of how we govern AI, rather than an implicit assumption that "AI" means the same thing across every product. It sets out the principles we hold ourselves to when we design AI features, and when we use AI tools ourselves in our own engineering process.
Scope
This policy applies to:
- AI features and AI-powered products that UIT designs, builds, or operates, including generative AI, conversational AI, copilots, and agentic systems such as those orchestrated through Quantum AgentOS.
- UIT's own internal use of AI tools in software engineering, code review, documentation, and operational work.
It does not extend authority over how a client organization uses a delivered product once it is deployed under that organization's own governance — clients remain responsible for the clinical, operational, and regulatory decisions they make with any software, including software UIT built for them.
Core Principles
Human oversight for clinical and consequential decisions
Any AI capability that touches a clinical, financial, or otherwise consequential decision is designed as decision support, not decision authority. A qualified person reviews, confirms, or can override the AI's output before it takes effect. AI narrows options, surfaces information, and drafts recommendations — it does not have the final word.
Transparency about AI interaction
Where a person is interacting with an AI system — a chat interface, a copilot, an automated message — that fact is disclosed. We do not design systems intended to make a user believe they are speaking with a human when they are not.
Bounded autonomy for agentic systems
Agents built on Quantum AgentOS or similar orchestration operate within a defined task scope: specific tools, specific data access, and specific actions they are permitted to take. When an agent encounters a situation outside that scope, or a threshold requiring judgment, it escalates to a human rather than improvising. Autonomy is scoped deliberately, not assumed by default.
Data minimization in training and inference
AI features are designed to use the minimum data required for the task — scoped inputs, retention limits on prompts and outputs, and de-identification where the task allows it. We do not treat "more data" as a default design choice for AI systems any more than for any other system handling sensitive information.
No autonomous clinical diagnosis or treatment decisions
No AI system UIT builds is designed to independently diagnose a condition, prescribe or alter treatment, or make a clinical determination without a licensed clinician in the loop. Where AI contributes to a clinical workflow, it is positioned as an aid to clinical judgment, not a replacement for it.
Ongoing monitoring and audit
AI system behavior is monitored and reviewed after deployment, not only validated once before launch. Where an AI feature underperforms, drifts, or produces output outside expected bounds, that is treated as an operational issue requiring a fix — the same discipline applied to any other production system.
What Our AI Systems Do Not Do
To be direct about the boundary, UIT's AI systems do not:
- Make unsupervised clinical diagnoses or treatment decisions.
- Replace the judgment of a licensed clinician, care manager, or other qualified professional.
- Act outside their defined operational scope without human approval.
- Represent themselves as human when disclosure of AI involvement is warranted.
- Operate without a path for a human to review, override, or stop their output.
This applies regardless of how capable a given model or agent is in a demo or evaluation setting. Capability is not the same as authority to act unsupervised, and we do not grant the latter based on the former.
Raising a Concern
If you believe an AI system built by UIT is behaving outside these principles, or you have a question about how a specific product uses AI, contact us directly at info@uitsoftware.com or through our contact page. We treat reports about AI behavior as we would any other product or security issue — reviewed and addressed, not dismissed.
Related Policies
This policy should be read alongside our Security & Compliance page, which covers the technical safeguards behind our platforms, and our Data Protection Policy, which covers how data is handled more broadly.
Questions about how
we build AI?
We're glad to walk through the specifics for your use case.