2026.11.01 ai security RAG Poisoning Attack: How One Document Rewrites Your Answers One crafted upload can fill your top-k, earn a citation and trigger a tool call; here is how it works and what actually stops it. read
2026.11.01 guides LLM Penetration Testing Scope: What to Put in the SOW Most LLM pentest budgets buy jailbreak screenshots; here is the layer-by-layer SOW that tests tools, retrieval, tenants and output sinks instead. read
2026.11.01 ai security MCP Server Security Checklist: 12 Checks Before You Connect Twelve checks that decide whether an MCP server may sit on an agent with private data and a way to send it out. read
2026.11.01 ai security Brazil Court Fines Lawyers Over Hidden Prompt Injection in Filing Lawyers hid white-on-white orders for the judge's AI in a petition, and a Brazilian labor court fined them for it. read
2026.10.31 guides OWASP Top 10 for LLM Applications: Real Examples Each OWASP LLM Top 10 risk as a pentest finding: the side effect, the blast radius, and the control that still holds on retest. read
2026.10.31 guides AI Agent Penetration Test Scope: A Checklist for Security Leads The checklist for scoping an AI agent pentest so the report answers what an attacker can make your agent do. read
2026.10.31 ai security RAG Security Testing: Seven Attacks on Retrieval Seven retrieval attacks you can prove with canaries, a second tenant, and document-ID traces, plus the architectural fixes that leave answer quality intact. read
2026.10.31 guides Penetration Test Cost in 2026: Web, API, Cloud and AI Agent Pricing What web, API, cloud, red team and AI agent penetration tests really cost in 2026, and how to spot a scan sold as a pentest. read
2026.10.31 guides API Penetration Testing Checklist for 2026: What Scanners Miss The API bugs that matter return HTTP 200, which is exactly why your scanner never flagged them; here is the checklist that finds them. read
2026.10.31 guides How to Test Excessive Agency in AI Agents Inventory every action an AI agent can reach, prove the chains in a sandbox, and score blast radius before you cut permissions. read