Vulnerabilities
Jev Model Flaw Lets Counterfeit Policies Bypass LangChain Security Gates
Research shows Jev’s classification logic can be steered by fake policies, allowing attackers to bypass tool-risk controls in applications.
prompt injection
Everything Patch Herald has reported about prompt injection: 2 stories, newest first. Part of our Vulnerabilities coverage.
Cyber Attacks
Barracuda: Emails now hide separate payloads for humans and AI to bypass review
Malwarebytes Labs warns that attackers embed secret instructions in emails to manipulate AI inbox tools into generating fraudulent to-do items for users.