ANSWER FIRST
Dated analysis for teams buying or running AI code security tools, written from public vendor pages, changelogs and standards. Each note lists its sources at the end.
Editorial assessment · Compiled from vendor product, docs, pricing and marketplace pages · Rated October 2026
Seven questions to ask before buying a coding agent security tool: assistant coverage, agent controls, secrets, package checks, prioritization and pricing.
A themed recap of AI code security announcements from June to September 2026: coding agent controls, malicious packages, secrets, fix agents, AI testing, and the CRA and MCP updates.
Published prices for AI code security tools, worked out for teams of 10, 25 and 100 developers, plus what the vendors without prices publish instead.
What the fix agents from Legit Security, Snyk, GitHub, Cycode, Aikido and Apiiro do, where they run and what they validate, from public vendor pages.
npm v12 stops running dependency install scripts by default and gates git and remote URL dependencies. What that changes for teams whose coding agents run npm install.
Five public documents that shape AI code security tools: OWASP's LLM Top 10, NIST SSDF, the EU Cyber Resilience Act, SBOM and the MCP specification, and what each asks for.
Six common assumptions about securing AI-generated code, from 'guardrails replace scanning' to 'npm v12 fixes malicious packages', checked against vendor pages and changelogs.
A plan for a proof of concept of coding agent security tools: scope, harmless test cases for the agent and the code, a scoring sheet and the questions a short test cannot answer.