The IIF has responded to the European Commission's consultation on guidelines for defining the scope of AI systems and prohibited practices under the EU AI Act with a short submission focusing on key questions.

In the submission, the IIF calls for a thoughtful approach that recognizes existing financial sector regulations and risk management frameworks and preserves beneficial AI applications. The submission emphasizes that the definition of AI systems should focus on adaptive and autonomous systems that can infer methods of generating results, and not capture traditional statistical and analytical methods that have been safely used for decades. It argues that traditional methods like linear regression, deterministic models, and supervised machine learning should be excluded from the Act's scope.

We stress that legitimate business practices like credit assessment, fraud prevention, anti-money laundering activities, customer segmentation, scanning of customer and trader calls, and adaptive employee-facing chatbots should be carved out from prohibitions. The submission also touches on issues including social scoring and emotion recognition in financial services. We also call for clarification on concepts like significant harm and autonomy to ensure practical implementation, while warning that a literal-minded approach to the guidance task will mean implementation quickly becomes unmanageable given the sheer number of software systems in everyday use in financial institutions.

The letter closes by offering to facilitate further discussions, including by sharing findings from our upcoming 2024 IIF-EY survey on AI in financial services.