{"data":{"id":"6bd0ad7a-3359-41f0-b556-12668c87ae68","title":"Prompt-Based Jailbreaking of Leading LLM Chatbots: A Survey of Attacks and Defenses","summary":"Large language models (LLMs, or AI systems trained on massive amounts of text) remain vulnerable to jailbreak attacks, which are adversarial prompts (tricky inputs designed to trick the AI) that bypass safety features and make the AI produce harmful or restricted content. This survey examines jailbreak techniques from 2023-2025, including prompt injection, role-playing tricks, and multi-turn conversations, alongside defense strategies like supervised fine-tuning (adjusting the model using labeled examples) and reinforcement learning from human feedback (improving the model based on human ratings of its outputs).","solution":"N/A -- no mitigation discussed in source.","labels":["security","research"],"sourceUrl":"http://ieeexplore.ieee.org/document/11397677","publishedAt":"2026-02-17T13:51:11.000Z","cveId":null,"cweIds":null,"cvssScore":null,"cvssSeverity":null,"severity":"info","attackType":["jailbreak","prompt_injection"],"issueType":"research","affectedPackages":null,"affectedVendors":["OpenAI","Anthropic","Google","Meta"],"affectedVendorsRaw":["OpenAI","Anthropic","Google","Meta","Claude","GPT","Gemini","Llama"],"classifierModel":"claude-haiku-4-5-20251001","classifierPromptVersion":"v3","cvssVector":null,"attackVector":null,"attackComplexity":null,"privilegesRequired":null,"userInteraction":null,"exploitMaturity":null,"epssScore":null,"patchAvailable":null,"disclosureDate":"2026-02-17T13:51:11.000Z","capecIds":null,"crossRefCount":0,"attackSophistication":"moderate","impactType":["safety","integrity"],"aiComponentTargeted":"model","llmSpecific":true,"classifierConfidence":0.95,"researchCategory":"peer_reviewed","atlasIds":null}}