Article 10: Data Governance Rules for AI — What You Need to Know

Why Data Quality Matters for AI

AI systems are only as good as the data they're trained on. If an AI hiring tool was trained mostly on data from one demographic group, it might not work fairly for everyone. If a medical AI was trained on incomplete patient data, its recommendations could be unreliable. Article 10 addresses this by setting standards for data quality.

For most small businesses, you're using AI tools built by others, not building your own. So the data governance burden falls primarily on your AI vendors. However, you still have responsibilities — particularly around the data you input into these tools and the questions you ask about your vendors' practices.

What Article 10 Covers

The article requires that training data be relevant, representative, free of errors (as far as possible), and complete. It also requires that data sets take into account the specific geographic, behavioral, or functional setting where the AI system will be used. For instance, an AI tool trained only on American English might not work well for a business serving customers who speak British English.

There are also requirements around bias detection. The data used to train high-risk AI must be examined for possible biases, especially concerning protected characteristics like race, gender, or age.

Your Responsibilities as a Business User

Even if you didn't build the AI tool, you should ask your vendor: what data was the system trained on? Has it been tested for bias? Is it appropriate for your specific use case and geographic context? These are reasonable questions that any responsible vendor should be able to answer.

You should also think about the data you feed into AI tools day-to-day. If you're using AI to analyze customer data, make sure that data is accurate and up-to-date. Garbage in, garbage out — no amount of regulatory compliance can fix AI outputs based on bad input data.

Practical Steps

Include data quality questions in your vendor assessment process. Review the data you regularly feed into AI tools for accuracy. Be cautious about using AI tools in contexts they weren't designed for — a tool trained on US legal data might give misleading results for UK legal questions. Document your data governance practices, even if they're simple, to show regulators you're taking this seriously.

This article is for informational purposes only and does not constitute legal advice. Regulatory requirements change frequently — verify current rules with official sources. Built by Sawai Gyoseishoshi Office, Hiroshima, Japan.