GPT-6 Astra Bypasses Safety Filters via Wrong Keyboard Layout
Researchers found GPT-6 Astra can evade content filters by typing in the wrong keyboard layout, exposing gaps in AI safety.
#prompt-injection×
NVIDIA Releases SkillSpector: Open-Source Agent Skill Security Tool
SkillSpector by NVIDIA is an open-source tool to find vulnerabilities in agent skills, including prompt injection, data leaks, unsafe code, and dependencies.