Overview
Severity: HIGH | Affected: Multiple LLM Providers | Category: research
A research team from Stanford University has published a paper detailing a novel jailbreak technique named 'Cognitive Override.' This method bypasses LLM safety protocols not through simple adversarial suffixes, but via complex, multi-turn conversations that gradually build a contradictory context. This manipulation leads the model into a confused state where it misinterprets its own safety guidelines, allowing it to respond to harmful prompts. The paper demonstrates a high success rate against several leading proprietary and open-source models, highlighting a fundamental vulnerability in current alignment strategies that rely on static, rule-based safety filters. The findings pressure AI developers to invest in more dynamic, context-aware safety mechanisms that can monitor the model's internal state throughout a conversation.