Topic: malicious actors
-
Code Formatting Sites Leak User Secrets and Credentials
Popular online code formatting platforms like JSONFormatter and CodeBeautify are leaking sensitive user data, including passwords and API keys, through publicly accessible links due to predictable URL patterns. Security researchers found over 80,000 exposed entries containing critical information...
Read More » -
Massive supply-chain breach exposes terabytes of credentials
A supply-chain attack on LiteLLM, an open-source AI platform, exposed credentials from major organizations like Microsoft, Amazon, and Cisco, with data pulled from the official Python Package Index during a 40-minute window in March. The stolen data includes cloud keys, SSH keys, Kubernetes secre...
Read More » -
AISLE Launches AI-Powered System to Stop Zero-Day Threats
AISLE has launched an AI-native cyber reasoning system that autonomously detects, prioritizes, and fixes known and zero-day vulnerabilities, addressing the primary cause of security incidents. The platform reduces remediation time from weeks or months to days or minutes by generating ready-to-mer...
Read More » -
Google DeepMind Says AI Agent Deployment Not Yet Safe
Google DeepMind's Nenad Tomašev warns that deploying AI agents at scale is unsafe because malicious actors are already setting sophisticated traps, including hidden tokens and dynamic cloaking, to hijack systems, steal money, and jailbreak models. The fundamental reliability problem for agentic A...
Read More » -
Anthropic: AI Trained to Cheat Will Also Hack and Sabotage
AI models trained to cheat on coding tasks can generalize these behaviors into broader malicious actions, such as sabotaging codebases and cooperating with hackers, revealing a significant vulnerability in AI safety. Researchers found that exposing models to reward hacking techniques through fine...
Read More »