AI Security — Cybersecurity News

Latest News

ChatGPT Lockdown Mode Bypassed: Researchers Demonstrate Silent Chat Exfiltration

ChatGPT Lockdown Mode Bypassed: Researchers Demonstrate Silent Chat Exfiltration

RedSide Security November 06, 2026 AI Security 166 views

Researchers found and disclosed a now-patched flaw that bypassed ChatGPT's Lockdown Mode, allowing silent chat data exfiltration without user interaction — highlighting the persistent challenge of prompt-injection defense.

Continue reading: ChatGPT Lockdown Mode Bypassed: Researchers Demons…
Google Warns LLMjacking Attacks Are Draining Cloud Budgets by Up to $1 Million

Google Warns LLMjacking Attacks Are Draining Cloud Budgets by Up to $1 Million

RedSide Security November 03, 2026 Cloud Security 83 views

Google is warning that attackers are stealing enterprise LLM API credentials to abuse paid AI compute, with some victims facing cloud bills approaching $1 million. The tactic, known as LLMjacking, exploits leaked keys, misconfigurations, and exposed client-side code.

Continue reading: Google Warns LLMjacking Attacks Are Draining Cloud…
Critical Langflow Flaws Allow Unauthenticated Code Execution on AI Workflow Servers

Critical Langflow Flaws Allow Unauthenticated Code Execution on AI Workflow Servers

RedSide Security November 02, 2026 Vulnerability 64 views

Two critical vulnerabilities in Langflow affect every release from 1.0.0 to 1.12.8, allowing unauthenticated attackers to bypass authentication or execute arbitrary code. Only version 1.13.0 contains the fix.

Continue reading: Critical Langflow Flaws Allow Unauthenticated Code…
Google Patches 10 Chrome Vulnerabilities, Including a Memory Corruption Flaw Surfaced With Anthropic's Claude AI

Google Patches 10 Chrome Vulnerabilities, Including a Memory Corruption Flaw Surfaced With Anthropic's Claude AI

RedSide Security October 17, 2026 Vulnerability 331 views

Google's latest Chrome Stable update fixes 10 vulnerabilities across Windows, macOS, and Linux, including a high-severity DOM memory corruption flaw reported by an Anthropic researcher with the help of Claude AI.

Continue reading: Google Patches 10 Chrome Vulnerabilities, Includin…
Insider Threat at OpenAI: Employee Charged With Planting Hidden Recorder in Executive Conference Room

Insider Threat at OpenAI: Employee Charged With Planting Hidden Recorder in Executive Conference Room

RedSide Security October 14, 2026 Cybersecurity 108 views

An OpenAI employee has been charged with allegedly planting a hidden recording device in an executive conference room in San Francisco, spotlighting the persistent risk of insider threats at AI companies.

Continue reading: Insider Threat at OpenAI: Employee Charged With Pl…
Encrypted AI Reasoning Traces Can Leak Secrets Across API Sessions

Encrypted AI Reasoning Traces Can Leak Secrets Across API Sessions

RedSide Security August 12, 2026 AI Security 315 views

A new study shows that encrypted AI reasoning traces from OpenAI, Anthropic, and Google could be replayed and decoded to expose hidden reasoning, API keys, passwords, and other sensitive data. Researchers also demonstrated hidden prompt injection attacks, highlighting risks in how proprietary LLM APIs handle encrypted reasoning objects across sessions and models.

Continue reading: Encrypted AI Reasoning Traces Can Leak Secrets Acr…
Critical Fast-MCP-Telegram Vulnerability Allows Authentication Bypass via Path Traversal

Critical Fast-MCP-Telegram Vulnerability Allows Authentication Bypass via Path Traversal

RedSide Security July 07, 2026 Vulnerability 294 views

A critical authentication bypass vulnerability (**CVE-2026-52830**) in Fast-MCP-Telegram allows attackers to exploit path traversal flaws in Bearer token validation and gain unauthorized access to Telegram sessions. Users are strongly advised to upgrade to version 0.19.1 immediately.

Continue reading: Critical Fast-MCP-Telegram Vulnerability Allows Au…
Critical Cursor IDE Flaws Enable Full Sandbox Escape and Remote Code Execution

Critical Cursor IDE Flaws Enable Full Sandbox Escape and Remote Code Execution

RedSide Security July 01, 2026 Vulnerability 268 views

Two critical vulnerabilities in Cursor IDE, tracked as CVE-2026-50548 and CVE-2026-50549, allow attackers to escape the platform's sandbox and achieve full remote code execution through prompt injection. The flaws demonstrate how AI-driven coding agents can expose traditional software attack surfaces, leading to system compromise without user approval.

Continue reading: Critical Cursor IDE Flaws Enable Full Sandbox Esca…