AI Models Overthink Problems—and It’s a Security Risk
IEEE Spectrum AI Edd Gent ● Covered by 2 sources
Researchers from Zhejiang University and Alibaba developed an evolutionary algorithm that generates logically inconsistent prompts to cause reasoning models to produce excessive internal monologues, creating a denial-of-service vulnerability. The attack produced outputs up to 26.1 times longer than normal responses on the MATH dataset and proved effective against models including DeepSeek-R1, Qwen3-Thinking, GPT-o3, and Gemini 2.5 Flash. The findings highlight a shared security weakness in modern reasoning-capable LLMs that providers should address through mitigation strategies.
Why it matters
Large language models (LLMs) that can think through problems step-by-step have significantly increased the scope of tasks that AI can tackle. But new research suggests these reasoning capabilities also introduce a critical vulnerability that could allow attackers to slow these systems to a crawl.While earlier generations of LLMs would immediately produce a response to a user’s request, today’s most advanced models generate an internal monologue where they break down the problem into steps and reason about the best way to tackle it before providing an answer. This has allowed AI to tackle increasingly complex problems, particularly in areas like coding and math.However, previous research has shown that these models are susceptible to sometimes producing excessively long streams of reasoning that do little to boost performance, a phenomenon known as “overthinking.” In research presented this week at the International Conference on Machine Learning 2026, in Seoul, researchers from Zhejian
