The Issue With Explainable AI (XAI)

← All articles
Artificial intelligence · EN

The Issue With Explainable AI (XAI)

The EU GDPR’s transparency requirement mandates that organizations must be clear about how they collect, store, and use personal data. The requirements are relatively straightforward. On the other hand, the AI Act’s explainability requirement goes a step further. According to the EU AI Act’s Recital 59 and Article 13, high-risk AI systems must be transparent, explainable, and documented. However, the Act does not provide a precise definition or tangible thresholds for what ‘explainable’ truly means.

This is where XAI comes in - a suite of methodologies that aim to make AI’s output more comprehensible and trustworthy for humans. However, a recent scientific paper raises some serious concerns I share. It suggests that the current use of XAI may be misleading, providing unreliable explanations and false impressions about the workings of AI systems.

When we use suche tools, we need to ensure they do not become a factor of inequality or exclusion. We need asap clear definitions and standards for explainability to foster trust and confidence in AI systems and to ensure accountability.

I welcome your thoughts.

#AI #DataProtection

This post has a document attached on LinkedIn. View on LinkedIn →

Discuss on LinkedIn