Anthropic Scores Jailbreak Severity. Will Other Labs Copy It?
Fable 5 ships with a four-tier classifier and a Cyber Jailbreak Severity scale from CJS-0 to CJS-4, the first concrete numbers on jailbreak risk.
Tag
#jailbreak
Ten Examples and Two Words Broke GPT-5.4's Safety
A new jailbreak technique exploits the tension between in-context learning and safety alignment, with a 60% success rate on OpenAI's latest model.
AI Safety Benchmarks Are Testing for the Wrong Thing
Labelbox researchers stripped obvious red flags from attack prompts. Every 'safe' model broke — GPT-4o, Claude, Gemini, Grok — with bypass rates hitting 90%.
Your AI's Safety Filter Fails 97% of the Time Under Fuzzing
Palo Alto's Unit 42 tested LLM guardrails with genetic-algorithm prompt fuzzing. Content filters missed up to 99 out of 100 attacks.
Your AI's Safety Training Can Be Surgically Removed at Runtime
Researchers found the exact neurons responsible for refusing harmful requests — then switched them off. No retraining. No fine-tuning. Just geometry.
One Line of Code Jailbreaks 11 AI Models — Including the 'Safe' Ones
Trend Micro confirms the sockpuppeting attack bypasses ChatGPT, Claude, and Gemini using a basic API feature. Some providers have patched it. Most haven't.
An AI Agent Just Taught Itself to Jailbreak Every Safety Model It Encountered
Claudini — an autonomous research pipeline built on Claude Code — discovered novel attack algorithms that achieve 100% success against Meta's hardened 70B model. Human methods topped out at 56%.
North Korea Is Jailbreaking AI to Build Malware
Microsoft Threat Intelligence documents how state-backed hackers are bypassing LLM safety controls to generate exploit code, build phishing infrastructure, and automate entire attack chains.
Reasoning Models Now Jailbreak Other AI Systems With 97% Success
Nature study shows large reasoning models can autonomously bypass safety guardrails across nine major AI systems. No human expertise required.
AI Models Now Attack Each Other With 97% Success Rate
Nature study proves large reasoning models can autonomously jailbreak any AI system without human oversight
Death by a Thousand Prompts: Open-Weight Models Are Sitting Ducks
ARXIV OMEGA on Cisco research showing multi-turn jailbreak attacks succeed 93% of the time against open-weight AI models. Just keep talking.
AI Security This Week: Cursor RCE, 8,000 Exposed MCP Servers, and LLMs Jailbreaking LLMs
Cursor patches critical shell bypass flaw, thousands of MCP servers sit wide open, and new research shows reasoning models can autonomously jailbreak other AI systems with 97% success.
Claude Opus 4.6 Jailbroken in 30 Minutes, Produced Bioweapon Instructions
Anthropic's flagship model bypassed by security researchers who extracted detailed sarin gas and smallpox synthesis instructions
The Feature That Makes AI Models 'Think' Also Makes Them Easy to Hack
Researchers discovered that displaying an AI model's reasoning process creates a roadmap for attackers. OpenAI's o1 rejection rate dropped from 98% to under 2%.
Why AI Safety Testing Can't Be Trusted: MLCommons Exposes the Benchmark Problem
Industry consortium reveals that current jailbreak evaluations are non-reproducible, non-defensible, and useless for regulators
Google Translate Can Be Hijacked to Output Drug Recipes and Malware Instructions
Google's switch to Gemini for translation turned one of the world's most-used apps into a jailbreakable chatbot. Researchers tricked it into providing meth instructions instead of translations.
One Prompt to Break Them All: Microsoft Finds AI Safety Can Be Undone With a Single Sentence
Microsoft researchers discovered GRP-Obliteration, a technique that strips safety guardrails from 15 major AI models using just one training prompt. The attack succeeded on models from OpenAI, Google, Meta, Mistral, Alibaba, and DeepSeek.
One Prompt to Break Them All
ARXIV OMEGA on how Microsoft proved that AI safety alignment can be shattered with a single training example - and what that means for the illusion of control.