Responsible AI Policy

Our Principles

We believe AI should be useful, safe and accountable to the people it affects. The commitments below guide how we design, build, deploy and support AI systems.

Human Oversight

People stay in charge of consequential decisions. When an AI system could significantly affect someone’s health, finances, employment, legal rights or access to services, we design for a human in the loop. This means a qualified person reviews, approves or can override the output before it takes effect.

We also build clear ways for people to pause, correct or turn off AI features. AI agents (software that can take actions on its own toward a goal) receive only the permissions their tasks require.

Privacy and Data Protection

We treat client data as the client’s data. Our commitments include:

• No training public models on client data. Client data is not used to train public AI models.
• Least-privilege access. People and systems get access only to the data they need.
• Data minimization. We use only the data a solution needs to do its job.
• Compliance-aware design. We work within your compliance program and design for requirements such as HIPAA where they apply. We support business associate agreements with model providers where required.

Security

AI systems bring new risks, such as prompt injection, data leakage and misuse of connected tools. We address these risks alongside standard security practices. That includes access controls, encryption where appropriate, secrets management, logging and security testing before release.

Our AI governance, security and compliance services help clients apply these same controls across their own AI programs.

Transparency

People should know when they are dealing with AI. Where appropriate, we design systems that tell end users they are interacting with AI or reading AI-generated content.

In retrieval-augmented generation (RAG) systems, the AI answers using your approved documents. We design these systems to cite their sources so users can check answers. We also document how systems work, what data they use and their known limits.

Fairness and Bias Testing

AI can repeat or amplify unfair patterns in data. Where a system could affect people differently based on protected characteristics, we test for bias. We review training data, evaluate outcomes across relevant groups and address problems we find. We share results and residual risks with the client.

Accuracy, Evaluation and Monitoring

AI outputs can be wrong. We evaluate systems against realistic test cases before launch and set clear quality thresholds with the client.

After launch, we recommend ongoing monitoring for accuracy, drift and unexpected behavior. When a system falls short, we investigate, adjust or roll it back.

Accountability

Every AI project has named people responsible for its design, testing and approval. We keep records of key decisions, evaluations and changes. Clients remain responsible for how they use AI in their operations, and we help them set up governance to do that well.

AI-Assisted Software Development Standards

Our engineers use AI tools such as coding assistants, automated test generation and AI code review. These tools help us deliver working software sooner. They do not replace engineering judgment. Learn more about our AI-accelerated development approach.

Our standards for AI-assisted development:

• Human review. Every AI-generated change is reviewed by experienced engineers before it is accepted.
• Testing. AI-generated code is tested like any other code.
• Security scanning. AI-generated code is security-scanned before release.
• Client ownership. Clients own their code and intellectual property, as set out in their agreements.
• Data protection. Client code and data are not used to train public AI models.

Third-Party Models and Vendors

Many AI solutions use models and services from third-party providers. We review providers’ data handling, security and terms of use before recommending them. Where possible, we choose configurations that keep client data out of provider model training.

We tell clients which third-party models and services a solution relies on. We monitor provider changes that could affect performance, cost or risk.

Alignment with Recognized Frameworks

Our practices are informed by recognized guidance, such as the NIST AI Risk Management Framework. We use this guidance to identify, measure and manage AI risks. NeoTek Solutions does not claim certification under this or any other AI framework.

Uses We Decline

We will not knowingly build or support AI systems intended for:

• Deceptive deepfakes or impersonation of real people without their consent
• Unlawful surveillance or tracking of individuals
• Discriminatory profiling or decisions based on protected characteristics
• Manipulating people in ways that cause harm
• Any purpose that violates applicable law

We may decline or end work that conflicts with this policy.

How to Raise a Concern

If you have a question or concern about how NeoTek Solutions builds or uses AI, please email info@neoteksol.com. We take concerns seriously and will review each one. You can also reach us through our contact page.

Review of This Policy

We review this policy regularly as AI technology, laws and good practice change. When we update it, we will change the “Last updated” date at the top of this page. For how we handle personal information, see our Privacy Policy.