AI Jailbreak & Security 95 posts

Safety research only. No attack recipes.

Sort:

AI Jailbreak & Security — All Posts

Prompt Guard 2’s Internal Representations Outperform Its Classification Head

JamieCrafterAdvanced· 383 · 4 · 8 ·8/23/2026 AI Jailbreak & Security AI Safety LLM Security

LLM security is a constant whack-a-mole battle against prompt injection

Zoe12Novice· 525 · 3 · 11 ·8/22/2026 AI Jailbreak & Security AI Safety LLM Security

LLM security is a constant whack-a-mole battle against prompt injection

Safety filters in multimodal LLMs fail when hidden prompts hide inside images

CyberSmithAdvanced· 602 · 4 · 11 ·8/21/2026 AI Jailbreak & Security AI Safety LLM Security

Dynamic learning outpaces fixed defenses in COPA’s evolving prompt attack countermeasure strategy.

Zoe12Novice· 698 · 3 · 8 ·8/21/2026 AI Jailbreak & Security AI Safety LLM Security

OpenAI safety monitoring now consumes one-fifth of inference FLOPs for active oversight

DeepSurferNovice· 552 · 3 · 7 ·8/20/2026 AI Jailbreak & Security AI Safety LLM Security

OpenAI safety monitoring now consumes one-fifth of inference FLOPs for active oversight

Fair‑ASR Reveals New Attack Rankings by Normalizing Target Calls

NightPandaExpert· 569 · 3 · 0 ·8/19/2026 AI Jailbreak & Security AI Safety LLM Security

Tripwire approach stops jailbreaks without reducing large language model usefulness

Sam46Advanced· 101 · 3 · 3 ·8/18/2026 AI Jailbreak & Security AI Safety LLM Security

RLHF alone cannot protect against autonomous agents

QuinnPilotNovice· 573 · 4 · 0 ·8/17/2026 AI Jailbreak & Security AI Safety LLM Security

Qwen 3.8 27B Strikes the Perfect Balance for On-Premise AI

Cameron9Advanced· 290 · 3 · 1 ·8/16/2026 AI Jailbreak & Security AI Safety LLM Security

Rethinking LLM Intelligence: It’s Not About Magic Words

NightPandaExpert· 128 · 4 · 0 ·8/15/2026 AI Jailbreak & Security AI Safety LLM Security

Pretraining alignment needs depth, not surface fixes, for fundamental model learning changes

KaiDevExpert· 548 · 4 · 9 ·8/14/2026 AI Jailbreak & Security AI Safety LLM Security

ProbGuard flags jailbreak attempts inside the first ten tokens

DrewCoderNovice· 355 · 3 · 15 ·8/13/2026 AI Jailbreak & Security AI Safety LLM Security

How encrypted reasoning in Frontier LLMs could be exposed by model family key overlaps

IndieFounderIntermediate· 576 · 4 · 1 ·8/12/2026 AI Jailbreak & Security AI Safety LLM Security

Claude Auto Mode Becomes Default for Paid Subscribers

CameronOwlExpert· 571 · 3 · 5 ·8/11/2026 AI Jailbreak & Security AI Safety LLM Security

Reframing Safety as Missingness Beats Random Mixing in LLM Fine-Tuning

Morgan80Advanced· 427 · 3 · 7 ·8/10/2026 AI Jailbreak & Security AI Safety LLM Security

‹1 2 3 4 5 6 7 ›

Email