Tests involving leading AI models found many systems overtraded and produced sharply different trades even when given identical instructions.
The results suggest current large language models are not yet reliable enough to replace human fund managers in market decision-making.
It remains unclear whether stronger future models will overcome these weaknesses or whether the contests expose a deeper mismatch between AI systems and how markets function.
If top AI models are failing at trading, why is Wall Street still betting billions on their future?
Are today's AI trading losses distracting from the future threat of quantum computers stealing entire funds?