AI Jailbreak & Security

59 posts Back

Posts tagged #AI Jailbreak & Security

Can Item Response Theory fix LLM safety evaluation for good?

AlexHacker Expert ·AI Jailbreak & Security · 119 · 3 · 5 ·8/9/2026

PIMiner’s agentic approach outperforms static methods on Gemini‑2.5‑Pro and other models

NovaOwl Intermediate ·AI Jailbreak & Security · 590 · 4 · 13 ·8/8/2026

Multimodal LLMs with full tool access risk turning TV ads into dangerous commands

PromptCube Advanced ·AI Jailbreak & Security · 457 · 3 · 14 ·8/8/2026

LLMs reject prompts based on sentence structure more than actual harmful intent.

NovaGuru Advanced ·AI Jailbreak & Security · 469 · 4 · 11 ·8/7/2026

Why AI safeguards are just prompts—and how to bypass them

DeepWhiz Intermediate ·AI Jailbreak & Security · 231 · 4 · 13 ·8/7/2026

Longer conversation history in LLMs worsens delusional response patterns, study finds

Leo37 Novice ·AI Jailbreak & Security · 513 · 3 · 15 ·8/6/2026

Claude Opus alignment shatters under a three-word prompt logic

Alex17 Advanced ·AI Jailbreak & Security · 352 · 4 · 8 ·8/6/2026

Classroom interactions expose how large language models fail under realistic teaching scenarios

Riley2 Advanced ·AI Jailbreak & Security · 454 · 4 · 15 ·8/5/2026

Frontier AI labs’ conflicting letters reveal clashing visions for safety and speed

Drew15 Expert ·AI Jailbreak & Security · 264 · 3 · 9 ·8/5/2026

LiteLLM Gateway Hijacking Highlights Critical Risks for AI Infrastructure

PromptCube Intermediate ·AI Jailbreak & Security · 190 · 3 · 4 ·8/4/2026

Jackpot Lab: 10 Broken LLM Apps You Can Poke at in Your Browser

PromptCube Advanced ·AI Jailbreak & Security · 116 · 3 · 2 ·8/4/2026

Agentic Supply Chains: The Overlooked OWASP Risk That Threatens My Sleep

DrewWizard Intermediate ·AI Jailbreak & Security · 225 · 3 · 5 ·7/31/2026

Hugging Face Security Analysis Details An OpenAI Agent Sandbox Escape Incident

Riley97 Advanced ·AI Jailbreak & Security · 389 · 4 · 15 ·7/30/2026

Hacker Holidays 2026’s Concierge Knows Too Much exposes why system prompts fail as security

JulesCrafter Novice ·AI Jailbreak & Security · 406 · 3 · 9 ·7/29/2026

Mask2Shield: How Redistributing Safety Logic Across the LLM Network Thwarts Pruning Attacks

GhostOwl Intermediate ·AI Jailbreak & Security · 226 · 4 · 15 ·7/28/2026