GEAB 199

ANTICIPATION

As artificial intelligence models become increasingly prevalent in risk management, finance is undergoing a transformation and prediction, once a factor of control, is becoming a source of instability. By 2030, regulators could entrust collective algorithmic architectures with the task of “stabilising” markets, at the cost of a loss of human understanding. This shift towards a “blind system” based on fuzzy logic would represent a profound paradigm transformation – with both positive and negative implications.

The ability to predict, quantify and “rate” risk is a fundamental pillar of the contemporary financial system. This ability governs prices, capital allocations, capital requirements, investor confidence and the stability of the system. From credit ratings to bank stress test models, the “risk model” has long been a statistical tool based on assumptions, historical data and safety margins.

The introduction of AI in this field is a technological breakthrough as AI, through machine learning, large language models (LLMs), deep neural networks ( deep neural networks), promises to capture weak signals, cross-reference massive data sources (markets, macroeconomics, social networks, geopolitics, climate, supply chains) and anticipate non-linear events.

But this promise also masks profound dangers, such as cognitive leverage, invisible biases, concentration and mechanical risks (model failures, automated synchronisation, “avalanches”) that can transform a local crisis into a systemic cascade.