OpenAI Launches ChatGPT Lockdown Mode to Block Prompt Injection Data Exfiltration
OpenAI has released ChatGPT Lockdown Mode, a new security feature that disables outbound network capabilities to cut off data exfiltration pathways exploited in prompt injection attacks. Available to...
Critical HuggingFace Transformers Flaw CVE-2026-4372 Enables Silent RCE — 232 Million Installs at Risk
A critical RCE vulnerability in HuggingFace Transformers (CVE-2026-4372) allows attackers to silently execute code by loading a malicious AI model, bypassing the trust_remote_code=False security control. Over 232 million...
Google Gemini Voice Assistant Hijacked via WhatsApp, Slack and SMS: Researchers Bypass All Google Defenses
SafeBreach researchers demonstrate how attackers can silently hijack Google Gemini through malicious payloads in WhatsApp, Slack, SMS, and other messaging app notifications, bypassing all of Google patched defenses...
Threat Actors Use AI Agents and Cursor IDE to Automate Active Directory Attacks and Beat EDR
Sophos has uncovered a Russian-speaking threat actor using AI-assisted tools, Cobalt Strike, and a purpose-built automated lab to develop EDR bypass malware targeting Active Directory environments — with...
Five OpenClaw Zero-Days Let Attackers Silently Hijack AI Agent Access on Slack, Teams, and Discord
Researcher Philip Garabandic disclosed five zero-day vulnerabilities in OpenClaw allowing identity spoofing to hijack trusted AI agent access across Slack, Discord, Microsoft Teams, Matrix, and Zalo. The same...
Meta AI Flaw Lets Attackers Hijack Instagram Accounts Without Verification — Premium Handles Worth $1M+ Stolen
A critical flaw in Meta's AI account recovery tool allowed attackers to trick the chatbot into sending password reset codes with no identity verification, enabling theft of premium...
BadHost (CVE-2026-48710): Critical Authentication Bypass Threatens Thousands of AI Agent Applications
A newly disclosed critical vulnerability dubbed 'BadHost' (CVE-2026-48710) enables attackers to bypass authentication in FastAPI and Starlette-based AI applications through manipulated HTTP Host headers. Thousands of LLM inference...
AI Discovers 10,000+ Zero-Days: Anthropic’s Claude Mythos Preview Transforms Cybersecurity Defense
Anthropic's Claude Mythos Preview AI model has autonomously discovered over 10,000 critical zero-day vulnerabilities across major software systems as part of Project Glasswing, revealing both the extraordinary potential...