ThreatDeck Weekly brief · 2026-W46 · Nov 9 – 15, 2026
The week in AI security
What needs action, the highest-signal stories by pillar, the techniques that kept coming up and what joined the tool radar.
2 AI-security stories: 1 on attacks against AI, 1 on AI used for attack, 0 on AI for defense.
No AI-related item was reported exploited or needing a patch.
20 new tools on the radar.
Attacks on AI 1
-
CVE-2026-125771: Privilege escalation in AtlasAiDev MCP Server for WordPress through 2.0.3
An Incorrect Privilege Assignment vulnerability exists in AtlasAiDev MCP Server for WordPress (ai-workflow-automation-ai-agent-hub).
Offensive AI 1
-
Anthropic's Claude Fable 5.2 doesn't generate images, it paints them
Anthropic's Claude Fable 5.2 reportedly generates images by writing Python code that places simulated brushstrokes on a virtual canvas rather than computing pixels.
New on the radar
- agent-safehouseAgent & MCP security
- AgentZAgent & MCP security
- aktoAgent & MCP security
- Argus-Agentic-SOC-PlatformAI SOC
- AugustusLLM red-teaming
- bernsteinAgent & MCP security
- BoxPwnrBenchmark
- CascoOffensive agent
- cc-safety-netAgent & MCP security
- clawsecAgent & MCP security
Security essentials
-
The Microsoft Outage Left 15 Hotel Guests Poisoned by Chlorine
A Microsoft 365 outage blocked email chlorine-level alerts at a German hotel sauna, reportedly leaving 15 guests poisoned by excess chlorine.
Generated by ThreatDeck from public sources. One-line summaries are AI-assisted; read the source before acting. Week 2026-W46, Nov 9 – 15, 2026 (UTC).