Safety · IEEE Spectrum AI ·
AI Models Overthink Problems—and It’s a Security Risk
Researchers reported that step-by-step reasoning in large language models can be exploited to trigger excessive internal deliberation. By feeding models logically inconsistent prompts, they induced overthinking that could slow commercial systems and create a denial-of-service risk.