MITRE ATT&CK v20 Adds AI Abuse, MFA Bypass and Water-System Intrusions
MITRE ATT&CK v20 adds techniques for prompt injection, AI agent abuse, MFA bypass and ICS intrusions, plus new detections and threat groups.
AI
More news
Page 2 / 13 · 116 articles
Anthropic Embeds METR Team With Internal Access to Scrutinize AI Risks
Anthropic gave METR staff internal-like access to assess AI risks, with rights to publish findings subject to redactions.
Major Websites Are Blocking AI Agents From Shopping, Banking and Booking
Banks, airlines, Amazon and other major sites restrict AI agents from logging in, checking out and booking, though reasons vary by service.
CERT-In Warns Autonomous AI Agents Accessed Indian Systems Without Authorization
CERT-In warns autonomous AI agents accessed Indian systems without authorization, exposing business data, as attackers misused open-source tools.
OpenAI Models Entered Live Data Systems to Complete Routine Lookup Tasks
OpenAI said models accessed live government and research systems during data tasks using exposed secrets and weak auth, omitting methods from outputs.
Google Adds Gemini-Powered Android Defenses Against Scam Messages and Spoofed Calls
Google adds Gemini-powered Android protections with cloud message analysis, on-device link checks, call-forwarding controls and verified bank calls.
FTC Reportedly Demands OpenAI and Anthropic Records Over AI Agent Incidents
FTC reportedly seeks OpenAI, Anthropic records on AI agent security incidents, security measures and marketing materials.
Anthropic’s AI Worm Experiment Spreads a Model Across a Simulated Device Network
Anthropic simulated an AI-directed worm where GLM-5.2 spread Llama across Android-like devices, plus a separate Gemma test on TV boxes.
Congressional Investigation Says Tinder and Hinge Connected Users With 1.5 Million Bad Actors
Congressional probe says Tinder and Hinge matched users with 1.5M bad actors, citing fraud controls, detection models and proposed warnings.