Responsible AI policy

1. Our principles

• People stay in charge: AI supports decisions; people remain accountable for them.
• We are open about AI: people should know when they are dealing with an AI system.
• We test before we ship and keep testing after: for accuracy, fairness, security and failure.
• We protect data: client data is used only for that client, as instructed.
• We follow the law in every country where a system is used.

2. What we will not build

We do not build or support AI systems that are designed to:

• monitor, track or identify people at scale without a clear lawful basis, such as indiscriminate surveillance of public spaces
• score or rank people on their social behaviour or personal characteristics in ways that lead to unfair treatment
• manipulate people through deception or by exploiting their vulnerabilities
• create deceptive content, such as deepfakes of real people, intended to mislead
• cause physical harm, including weapons systems

3. Higher-risk systems

Some systems need extra care because mistakes can seriously affect people: for example systems that use biometric data (such as faces), or that inform decisions about credit, jobs, health, public services or legal rights. For these we only proceed when:

• there is a lawful basis for the processing, and the client confirms it in writing
• a data protection and risk assessment has been done before development starts
• a person reviews any decision with legal or similarly significant effects on someone, and people can challenge it
• the system has been tested for accuracy and for unfair differences between groups of people, and the results are shared with the client

4. Human oversight

Every system we deliver has a named owner on the client side, clear limits on what it may do on its own, a way for people to override or switch it off, and logs that show what it did. AI agents that can take actions, such as sending messages or changing records, work within permissions the client sets and approves.

5. Testing and monitoring

• Before launch: we test accuracy on data that reflects the people and languages the system will serve, check for bias, and test security, including prompt injection and data leakage for systems built on large language models.
• After launch: we agree with the client how performance is monitored, what counts as a failure, and who responds.
• We document what each model is for, its known limits and the data it was trained or evaluated on.

6. Client data and model training

Each client decides what data its system may use, and we process it only on the client's written instructions. We do not use one client's data to train models for anyone else. When we use third-party AI models or services, we choose options that do not use client data for the provider's own training wherever they are available, and we tell the client which providers are involved.

7. Transparency

We design systems so that people know when they are interacting with AI, and so that AI-generated content can be identified where that matters. We give clients plain-language documentation they can share with their own users.

8. Reporting a concern

If you think an AI system we built has caused harm, made an unfair decision or is being misused, email info@codersclutch.com with "Responsible AI" in the subject line. We will look into it, work with the client concerned and tell you what we found.

9. Keeping this policy up to date

Our leadership team reviews this policy at least once a year, and whenever laws or the technology change significantly.