ALL POSTS

Key Findings Check Point discovered a critical vulnerability in ChatGPT that allowed attackers to exfiltrate user data, uploaded files, and conversation history without detection or consent The flaw exploited a hidden DNS-based communication channel in the Linux runtime environment, bypassing all visible AI guardrails OpenAI patched the ChatGPT vulnerability on February 20, 2026, with no evidence of malicious exploitation BeyondTrust Phantom Labs identified a command injectio

Key Findings GlassWorm campaign evolved to deliver multi-stage malware framework with data theft and remote access capabilities Operators use Solana blockchain transactions as dead drop resolvers to hide command-and-control infrastructure Malware includes hardware wallet phishing targeting Ledger and Trezor devices with fake recovery phrase prompts Chrome extension masquerading as "Google Docs Offline" steals browser data, cookies, and monitors cryptocurrency exchange session

Key Findings * Critical remote code execution vulnerability in Langflow (CVE-2026-33017) * CVSS score: 9.3 * Exploited within 20 hours of advisory publication * Allows unauthenticated remote code execution via API endpoint * Affects all Langflow versions prior to 1.8.1 * Attackers can execute arbitrary Python code with full server privileges * Observed exploitation includes credential harvesting and potential supply chain compromise Background Langflow, an open-source AI plat

Key Findings DarkSword is a sophisticated iOS exploit kit targeting devices running iOS 18.4-18.7 Developed by UNC6353, likely a Russia-linked group Exploits six vulnerabilities, including three zero-days Enables full device compromise with minimal user interaction Targets sensitive data, including credentials and crypto wallet information Operates in a "hit-and-run" approach, exfiltrating data quickly and then cleaning traces Background DarkSword emerged in late 2025 as a po

Key Findings * Researchers discovered a vulnerability in AWS Bedrock AgentCore Code Interpreter * DNS queries can be exploited to leak sensitive data from supposedly isolated AI systems * Vulnerability received a high-risk severity score of 7.5/10 * AWS responded by updating documentation instead of creating a full patch * Potential risks include data breaches and infrastructure compromise Background AWS Bedrock is a platform for building AI applications, with the AgentCore C

Key Findings * GlassWorm malware campaign targeting Python repositories * Attackers use stolen GitHub tokens to force-push malicious code * Targets Python projects including Django apps, ML code, and PyPI packages * Earliest injections traced to March 8, 2026 * Uses a new offshoot called "ForceMemo" * Leverages malicious VS Code and Cursor extensions to steal credentials * Payload includes cryptocurrency theft and data exfiltration capabilities Background The GlassWorm attack

Here's the markdown-formatted article based on the source material: Key Findings OpenClaw AI agent has multiple critical security vulnerabilities Prompt injection attacks can lead to data exfiltration and unauthorized system access Chinese authorities have moved to restrict OpenClaw usage in government and military environments Malicious actors are exploiting the platform's popularity to distribute malware Background OpenClaw is an open-source, self-hosted autonomous AI agent

Key Findings * UNC6426 breached a victim's cloud environment within 72 hours * Supply chain attack compromised nx npm package in August 2025 * Stolen GitHub token used to gain unauthorized cloud access * Threat actor created new AWS administrator role * Exfiltrated data from S3 buckets and destroyed production environments * AI-assisted attack leveraged LLM tools for credential theft Background The incident originated from a supply chain vulnerability in the nx npm package di

Key Findings Cybersecurity researchers have uncovered a new information stealer that exfiltrated a victim's OpenClaw configuration environment. The incident marks a significant evolution in infostealer behavior, transitioning from stealing browser credentials to targeting the identities, settings, and "digital souls" of personal AI agents. The stolen files included openclaw.json with gateway tokens, device.json containing private cryptographic keys, and "soul" and memory file

Key Findings Ivanti released security patches for its Endpoint Manager (EPM) product, addressing two critical vulnerabilities. The most severe flaw, CVE-2026-1603, is a high-severity authentication bypass (CVSS 8.6) that allows remote unauthenticated attackers to access stored credentials. The second vulnerability, CVE-2026-1602, is a medium-severity SQL injection flaw (CVSS 6.5) that could enable data theft by authenticated attackers. There is no evidence of these vulnerabil

Key Findings: A massive network of compromised YouTube accounts is being weaponized to spread a sophisticated new threat, turning the popular video platform into a launchpad for data theft. The campaign, dubbed the "YouTube Ghost Network," leverages malicious videos promoting "cracked" software, trainers, or cheats to lure users into downloading a new, heavily obfuscated JavaScript malware loader called GachiLoader. GachiLoader is written in Node.js and deploys a second-stage

Key Findings Over 30 security vulnerabilities have been disclosed in various AI-powered Integrated Development Environments (IDEs) The vulnerabilities combine prompt injection primitives with legitimate IDE features to achieve data exfiltration and remote code execution The security issues have been collectively named "IDEsaster" by security researcher Ari Marzouk (MaccariTA) The vulnerabilities affect popular IDEs and extensions such as Cursor, Windsurf, Kiro.dev, GitHub Cop