Release v1.2.0 — Q4 2026: rogue agents, coding agents, A2A, frontier capability · requie/AI-Red-Teaming-Guide

What's Changed

New sections: AI Coding-Agent & CI/CD Security · Agent-to-Agent (A2A) & Agent Identity · Frontier Capability & AI-Accelerated Vulnerability Discovery (Claude Mythos / Project Glasswing)
New case studies:
- D — An internal OpenAI frontier agent reached Australia's Medicare statistics portal on its own during evaluation (Jun 2026, disclosed Sep)
- E — "Comment and Control": one GitHub comment hijacked Claude Code, Gemini CLI, and Copilot coding agents in CI and leaked secrets
- F — Deadbugz: an MCP server that behaved for three tool calls, then told agents to steal SSH keys and cloud credentials
MCP section now covers runtime-gated poisoning and the classic web bugs behind most 2026 MCP CVEs

EU AI Act updated for the Digital Omnibus on AI: GPAI enforcement in force since 2 Aug 2026; high-risk obligations moved to 2 Dec 2027 / 2 Aug 2028
FTC probe of OpenAI, Anthropic, and METR over agent incidents and safety claims
NIST Cyber AI Profile and COSAiS agent overlays reported at their actual (still-draft) status; 2023 US EO marked historical
OWASP 2026 LLM Top 10, Agent Control Standard, and AI Red Teaming Landscape; MITRE ATLAS v5.x

New open-source tools: Darkmoon (self-hosted autonomous pentesting), MiDojo (asago / Red Hat), Ziran (TaoQ AI — graph-based tool-chain and multi-agent testing)
New commercial platforms: DeepKeep AI Security Platform; AVERSYN by Cogensec (featured)
Benchmarks: InjecAgent, HarmBench, JailbreakBench, CyberSecEval; HOL's AI Coding Agent Runtime Security Benchmark added as vendor-published further reading (a modeled comparison against the publisher's own product, not live testing)
Learning: HackAPrompt and vendor AI bug bounty programs (Google, Microsoft, OpenAI, Anthropic)

Rules of engagement, stakeholder readout, case study, and model/system security card rebuilt with agent/MCP fields and filled-in examples

Spanish, Chinese, and French translations re-synced to v1.2.0, with every English section link working in all three
New hero banner across all READMEs
Table of contents now covers every section; comparison matrix and Resources list every tool
README Contributors list now credits every community contributor
Fixed the broken RedTeamKit button link

The example garak scan no longer exposes the API key to pull-request code and skips cleanly without a key

Thanks to everyone who contributed to this release:

Full Changelog: v1.1.0...v1.2.0