AI & LLM
AI and LLM covers generative models, agents, prompt injection, data security and new artificial intelligence tools. The collection connects capabilities, limitations, operational risks and their impact on technical work.
PoC Zealot: Autonomous AI Executes End-to-End GCP Cloud Attack
Unit 42’s Zealot project demonstrates how multi-agent AI systems can autonomously chain SSRF, credential theft, and BigQuery exfiltrat…
18 Malicious AI Extensions Exposed: Unit 42 Details Email Spying and RAT Risks
Palo Alto Networks Unit 42 has uncovered 18 AI browser extensions that masquerade as productivity tools while deploying RATs and spyin…
Chrome Internal Bug Reports Surge to 200+ as Google Leans on AI
Google addressed more than 200 internally discovered vulnerabilities in Chrome between March and May 2026. The spike aligns with the c…
Frontier AI: The Shift from Coding Assistant to Autonomous Threat Agent
Research from Unit 42 reveals that frontier AI models now possess the autonomous reasoning capabilities of full-spectrum security rese…
Microsoft Open-Sources RAMPART and Clarity to Secure AI Agent Workflows
Microsoft has unveiled two open-source security tools for AI agents: RAMPART, a Pytest-native framework for build-time testing, and Cl…
Trust3 AI Launches MCP Security: A Hardened Control Plane or Just Another Promise?
Trust3 AI has announced MCP Security to protect enterprise agentic workloads, focusing on connection verification, isolated tokens, an…
AI-Driven Mobile Attacks Hit New Record: Apps Compromised Within Two Hours of Release
The Digital.ai 2026 App Security Threat Report reveals that 87% of client-facing applications are now under systematic attack, with th…
1Password and OpenAI Partner to Provide Just-in-Time Credentials for AI Agents
1Password integrates its Environments MCP Server into OpenAI's Codex, enabling just-in-time credentialing for AI coding agents to prev…
AI-Powered Honeypots: Cisco Talos Flips the Script on Automated Threats
On April 29, Cisco Talos Intelligence researchers released a proof-of-concept aimed at neutralizing offensive asymmetry in cyberspace.…
AI Agents in Production: Addressing the Confused-Deputy Threat in Operational Automation
New research identifies a critical architectural gap in operational AI agents where a lack of separation between reasoning and executi…
AI Productivity Facade: 18 Malicious Extensions Discovered with RAT and MitM Capabilities
Palo Alto Networks’ Unit 42 has identified 18 high-risk AI browser extensions that surveil emails, steal prompts, and compromise user…
Zealot: How Autonomous AI Orchestrates Multi-Stage Cloud Compromise
Palo Alto Networks’ Unit 42 has demonstrated Zealot, a multi-agent PoC capable of executing end-to-end cloud attack chains without hum…