Code of conduct for AI development

A code of conduct for AI development is a set of written principles and guidelines designed to guide the behavior of developers, researchers, and organizations building artificial intelligence systems.

It outlines the ethical boundaries, safety expectations, and accountability standards that must be followed during the entire AI lifecycle—from design to deployment.

This matters because without shared norms and responsibilities, AI development can quickly spiral into harmful, opaque, and biased practices.

A clear code of conduct supports trustworthy systems, minimizes risks, and helps meet legal obligations such as those outlined in the [EU AI Act](https://artificialintelligenceact.eu/), ISO 42001, or the OECD AI Principles.

“Only 16% of AI professionals believe their organization has a clear ethical framework for AI development.”(Source: 2023 IBM Global AI Adoption Index)

What is a code of conduct for AI?

A code of conduct for AI defines how people and organizations should act when designing, training, testing, and using AI systems. It reflects shared values like transparency, fairness, accountability, and privacy. These codes often align with international frameworks or organizational missions.

Why codes of conduct matter for AI teams

They provide direction when legal requirements are unclear or outdated. AI governance and compliance teams rely on them to flag early risks, audit processes, and define what “responsible AI” actually means inside the company. Risk teams use them to align development choices with business values and public expectations.

A real-world example comes from the UK Information Commissioner’s Office (ICO). Its “AI auditing framework” includes expectations about fairness, accountability, and accuracy that companies must meet to avoid legal penalties. A code of conduct translates such expectations into daily practice.

Common principles in AI conduct codes

Many AI conduct codes around the world include the following principles:

• Transparency: Clear documentation of how the model was trained, what data it used, and what decisions it influences.

• Accountability: Assigning people or teams who are responsible for the model’s impact and updates.

• Non-discrimination: Ensuring data and models do not reinforce social or historical biases.

• Privacy: Applying techniques such as data minimization and differential privacy to protect user information.

• Human oversight: Keeping people involved in important decisions, especially those affecting rights or freedoms.

Best practices for implementing AI codes of conduct

Good principles are not enough—they must be applied properly. Here’s how organizations make them real.

Start by linking the code to clear roles and responsibilities. Developers should know what “compliance” looks like. Risk managers should have access to checklists or workflows that reflect the code. And executives should back it up with resources.

Best practices include:

• Build the code with your team: Co-create it with input from engineers, product managers, lawyers, and ethicists. This improves adoption.

• Train regularly: New hires, contractors, and partners should all learn what the code means and how to apply it.

• Review and revise: Every year, update the code based on new laws, technologies, or incidents.

• Audit compliance: Use tools to track whether real development work aligns with the stated principles.

Practical use-cases

• A chatbot company writes a code of conduct to prevent its AI from generating harmful or false medical advice.

• A financial institution updates its AI risk policy to include “non-discrimination” testing on all credit-scoring models.

• A social media firm creates guidelines to ensure AI moderation doesn’t remove lawful speech disproportionately from marginalized groups.

FAQ

What is the difference between a code of conduct and AI policy?

A code of conduct explains the values and behaviors expected during AI work. An AI policy includes formal rules, procedures, and legal terms that enforce those behaviors. The code often supports the policy.

Is a code of conduct legally binding?

Not on its own. But it can support legal compliance by guiding ethical decision-making and showing a proactive approach to regulation.

Who should write the code of conduct?

Ideally, a cross-functional team. Include engineering, compliance, legal, and ethics roles. Organizations can also consult civil society, impacted communities, or academic experts.

How often should the code be updated?

Annually is recommended, or sooner if there’s a major legal change, audit failure, or public incident involving your AI.