DepthThe conclusion and your first steps
Direct answer
How accurate and robust does our AI system have to be?
You describe: You want to know what performance level the regulation demands of a high-risk system, and how to demonstrate it when no statutory minimum percentage exists. Likely role: provider (you place the system on the market).
This applies now
- For this situation, the preparation phase matters most right now.
Coming up
- Article 15: accuracy, robustness and cybersecurityfrom 2 December 2027
- Article 9: risk management systemfrom 2 December 2027
- Article 13: transparency towards deployersfrom 2 December 2027
The regulation names no fixed percentage. It requires a level appropriate to the intended purpose, which you substantiate yourself, state in the instructions for use and maintain throughout the lifecycle, including resilience against errors, faults and attempts to manipulate the system.
Your first actions
- Set and test performance and security levels. Determine appropriate accuracy, test robustness against errors and misuse, and take AI-specific security measures.
- Set up an iterative risk management process. Identify and analyse known and reasonably foreseeable risks, evaluate them and take measures, repeating the cycle on every change.
- Provide complete instructions for use. Describe capabilities, limitations, accuracy, oversight measures and expected lifetime in comprehensible form.
General interpretation, not legal advice. Checked against Regulation (EU) 2024/1689 and the Digital Omnibus (EU) 2026/1744; the official source remains authoritative.
Full map for your situation