High-signal AI/security/automation notes.
A new paper analyzes Model Context Protocol (MCP) and argues key weaknesses are architectural: capability attestation, origin authentication for sampling, and trust propagation in multi-server setups.
CVE-2026-0621: a ReDoS bug in Anthropic’s MCP TypeScript SDK UriTemplate matching can let crafted URIs peg CPU and hang Node-based MCP servers.
Microsoft describes a runtime defense model for Copilot Studio agents: treat every tool invocation as a high-risk event, inspect context via webhook, and allow/block in real time.
Microsoft shipped emergency out-of-band updates for an actively exploited Office security feature bypass (CVE-2026-21509) involving OLE mitigations; CISA added it to KEV.
Microsoft outlines an AI-assisted workflow that extracts TTPs from reports, maps to MITRE ATT&CK, and does coverage/gap analysis against a detection catalog (with human validation).
Two sandbox escape vulnerabilities in n8n (CVE-2026-1470, CVE-2026-0863) can enable remote code execution on self-hosted deployments; what changed, why it matters, and what to do.
NIST’s CAISI opened an RFI asking for concrete practices and measurement methods to secure AI agent systems (incl. hijacking/backdoors) with comments due March 9, 2026.
OpenAI describes an automated red-teaming loop for browser agents (ChatGPT Atlas), using RL-trained attackers + simulation to discover prompt injections and ship adversarial training + safeguards.
A new Systematization of Knowledge paper surveys prompt injection against agentic coding assistants and MCP ecosystems, arguing for architectural mitigations over prompt filtering.
Varonis Threat Labs describes Reprompt: a one-click Copilot Personal deep-link technique (q=) combined with multi-request chaining that enabled silent data exfiltration; Microsoft says it has been patched.
Varonis disclosed ‘Reprompt’: a one-click deep-link prompt injection flow that could chain requests inside Copilot Personal to exfiltrate user data; Microsoft says it’s patched.
A GitHub Advisory (CVE-2026-22807) says vLLM could execute attacker-controlled code during model resolution via Hugging Face auto_map, even when trust_remote_code is false.
A new vm2 sandbox escape (CVE-2026-22709) allows untrusted JavaScript to break out via Promise callback sanitization bypass. Upgrade guidance + what to audit.
A practical comparison of Zapier, Make, and n8n for AI workflows: pricing, flexibility, reliability, and when to pick each.
A practical workflow to triage email: classify, summarize, propose a reply, and route the message automatically.
A practical way to track AI-adjacent CVEs (RAG stacks, vector DBs, inference servers, agent tooling), prioritize risk, and patch fast.
A short digest focused on AI-adjacent security risk: where CVEs show up in modern AI stacks and what to do about it.
Five practical Zapier+LLM automations you can set up quickly, with setup steps and testing checklist.
A security-first reading of the joint guidance on deploying AI systems securely: what it recommends, why it matters, and quick actions teams can take.