Skip to main content

Tag: jailbreak

5 articles

Researchers work in a modern lab with a laptop in the foreground.

Researchers Uncover Adware Hidden in OpenAI Codex Jailbreak Toolkit

Security researchers just uncovered a sneaky adware hidden in a popular OpenAI Codex Jailbreak toolkit, cleverly designed to inject unwanted ads into AI responses. This cleverly crafted malware, dubbed Haiou, alters configuration files to promote a shady third-party site.

Analyst 207
Cluttered office desk with laptop and monitor, blurred cityscape through window.

AI Agent Exploits Exposed Models in Autonomous Cyber Attacks

Symantec's Threat Hunter Team caught a sophisticated AI attack agent in the wild, designed to autonomously scan the internet for vulnerable endpoints and exploit them without human intervention. This autonomous tool was built using off-the-shelf jailbreak and exploitation techniques, and could run unattended for up to 100,000 steps.

Analyst 207
Cramped, dimly lit room with laptop on dusty desk surrounded by old computer equipment and wires.

Russian Hacker Exploits Claude AI in Commercial Pentest Platform

A Russian hacker spent just $4 on a grey-market Claude API key, and within months, was selling a powerful commercial pentest tool built using jailbreak techniques to exploit the AI. This all started with a detailed tutorial on March 31, where six clever methods were shared to bypass Claude's safety filters.

Analyst 207
A computer workstation with a laptop and scattered papers on a minimalist desk in a bright, neutral-colored room.

Anthropic's Fable 5 Model Quickly Jailbroken

Anthropic's supposedly secure Fable 5 model was quickly exploited, with its guardrails designed to prevent cyberattacks bypassed in just days. This rapid jailbreak raises concerns about the model's safety and reliability.

Analyst 207
Researcher sits at desk with laptop, looking thoughtful near window.

Researcher Uncovers Feds' Swift Response to Simple Code Prompt

A researcher discovered that a simple code prompt to "fix this code" sparked a surprisingly swift and alarmed reaction from federal authorities regarding the Fable 5 model. However, the researcher disputes that this incident constituted a jailbreak, downplaying its significance.

Analyst 207