AI deep dive: LLM jailbreaking
bugcrowd.com | research | #ai-security | #prompt-injection | #jailbreak | #llm | #chatgpt | #threats
Summary
How attackers jailbreak LLMs to leak data, generate misaligned content, or degrade performance, covering prompt-based jailbreaking techniques and defenses to protect AI infrastructure.
- Published
- Collected
--- id: 699 title: "AI deep dive: LLM jailbreaking" source_domain: bugcrowd.com source_url: "https://www.bugcrowd.com/blog/ai-deep-dive-llm-jailbreaking/" published_at: "2024-11-19 13:00" tags: ["ai-security", "prompt-injection", "jailbreak", "llm", "chatgpt", "threats"] phying_url: "https://phying.com/item/699" lang: en --- # AI deep dive: LLM jailbreaking ## Summary How attackers jailbreak LLMs to leak data, generate misaligned content, or degrade performance, covering prompt-based jailbreaking techniques and defenses to protect AI infrastructure. ## Intelligence Metadata - **Sources:** bugcrowd.com - **Published:** 2024-11-19 13:00 - **Original:** https://www.bugcrowd.com/blog/ai-deep-dive-llm-jailbreaking/ - **Phying News:** https://phying.com/item/699