Why Cloudflare's GLM-5.3 KV Cache Leaked Across Tenants
Cloudflare published a postmortem on 2026-10-26 titled “Cross-request KV cache exposure in the GLM-5.3 Workers AI model, and our remediation with …
Read more →#Security
59 posts
Solo.io's agentshield Watches Your Agent Perimeter for Drift
I wrote about agent containment four days ago: default-deny egress, no ambient credentials, an audit log, one agentgateway config. All of that is a perimeter …
Read more →AI Agent Containment: Why a Sandbox Alone Isn't Enough
A CSO Online piece built around an October 2026 Hacker News thread made a direct claim: agents aren’t going rogue , frontier labs are skipping basic …
Read more →Free Plumbing, Paid in Footage
There’s a new deal for homeowners in three cities. Your plumber is free. The price is footage. I’ve spent this series walking the connected house …
Read more →OpenAI Red-Teamed Its Sandboxes and Found a CPU Side Channel
OpenAI spent hundreds of billions of tokens trying to break its own agent sandboxes, and its agents won often enough to patch a Kata Containers CVE, a gVisor …
Read more →OpenAI's Patch the Planet: What Safeguards Agent Scanning Needs
OpenAI’s “Patch the Planet” initiative spent two months coordinating a fix for two critical PHP flaws, collectively called PHP2Shell, before …
Read more →Who Signs the Test Report for a Robot?
Every product I ever tested ended in the same place: a test report with a signature on it. The signature mattered more than the data. It said someone qualified …
Read more →include_content=False Didn't Stop pydantic-ai's MCP Token Leak
pydantic-ai shipped two security fixes on 2026-10-19. The one I reproduced, GHSA-xv4x-g8mx-ch56 , is a two-flag problem: include_content=False does exactly what …
Read more →What Is an MCP Gateway? agentgateway Examples
An MCP gateway is a proxy between AI clients and MCP servers, so agents reach many tool servers through one endpoint and every tool list and tool call passes …
Read more →The Robot Is the Next Node in Your House Cluster
In the first post of my house series, I asked you to count the connected things in your house . Then we walked through it room by room . TVs, routers, cameras, …
Read more →Stopping Agent Proxy Laundering at the MCP Gateway
On 2026-10-05 the Wikimedia Foundation disclosed that a handful of wiki edits it attributes to OpenAI-operated agents changed the configuration of a citation …
Read more →Two Grok Bots Leaked a CEO's Bank Balance to Slack
Two “separate” Grok Bots leaked a CEO’s bank balance into a public Slack channel on October 19. Nobody jailbroke anything. Nobody injected a …
Read more →Who Gave the Agent a Badge?
Walk into any office building and you get a badge. It opens the lobby. Maybe your floor. It doesn’t open the server room, the CFO’s office or the …
Read more →LiteLLM 1.107.0: Why Image Signing Didn't Stop the Backdoor
LiteLLM 1.107.0 landed on PyPI and Docker Hub on October 21, 2026, with a credential stealer built into every wheel, and it was cryptographically signed through …
Read more →Find Shadow MCP Servers in Your Environment
Find shadow MCP servers by checking the places they hide: the network and your clusters, where an MCP server answers a server/discover or initialize request; …
Read more →LiteLLM security record: a due-diligence timeline
LiteLLM is the easiest way to put one API in front of every model provider. That ease is why it ends up holding a company’s provider keys, virtual keys …
Read more →MCP Tool-Call Guardrails: Gating Agents Before the Tool Runs
agentgateway can now reject or rewrite an MCP tools/call request before the tool runs, and mask the result before the client sees it, from CEL expressions in …
Read more →agentgateway Guardrails How-To
Agents and apps already send prompts through a gateway. Content safety belongs on that hop: inspect the request before it reaches a model, and inspect the …
Read more →The Path-Validation Bug That Let Copilot Register a Rogue MCP Server
CVE-2026-103572 , disclosed by embracethered.com and patched by Microsoft on 2026-10-13, is a Remote Code Execution bug in Visual Studio and SSMS Copilot, CVSS …
Read more →MCP Audit Tap: See What a Third-Party Server Actually Sends
A Show HN post for a “WhatsApp MCP server running free on Cloudflare” hit 124 points and 69 comments in under 16 hours on October 19, and none of …
Read more →Room by room: what a self-copying AI could do in your house
Last week I asked you to count the connected things in your house . Parks Associates says the average US internet household has about 17 (source ). This week we …
Read more →agentgateway as an Egress Proxy
AI agents need outbound access to models, tools, registries, and APIs. An unrestricted sandbox will use every channel it can find. The durable pattern is a …
Read more →One Fix, Four Hundred Repos
Think back to December 2021. Log4Shell. If you were anywhere near security or platform work, you remember that week. The fix sounded simple: upgrade log4j. Then …
Read more →agentgateway Standalone: Guardrails, Budgets, Tool Governance
themsquared/agentgateway-standalone-demo runs Solo Enterprise for agentgateway as a single docker compose up, no Kubernetes cluster and no cloud provider …
Read more →What Is Shadow MCP? A Plain Definition
Shadow MCP is any MCP server your AI tools can reach that your security or platform team doesn’t know about, didn’t approve, and can’t see. …
Read more →Stopping a Filesystem MCP Server From Reading .env Files
Disclosure: I work at Solo.io, which created agentgateway and sells Solo Enterprise for agentgateway. The demos I cite are public repositories. A filesystem MCP …
Read more →The house is full of computers now. Anthropic just showed what an AI could do with them.
Count the connected things in your house. The TV. The doorbell. The robot vacuum. The thermostat. The router blinking in the closet. Parks Associates puts the …
Read more →How AI Agents Sharing a Sandbox Leak Each Other's Data
On 2026-10-10, Cybernews reported that Shane Mac, CEO of XMTP Labs, gave a Grok Bot agent read-only access to his personal bank account as a finance assistant. …
Read more →Prisma AIRS Guardrails for agentgateway in 150 Lines of Python
Plenty of security teams already run Prisma AIRS , Palo Alto’s AI runtime security service, and they want every prompt and LLM response checked by it. I …
Read more →Is Jev Actually Calibrated? The Reliability Curve on 240 Cases
TypeSafe AI’s Jev is mostly calibrated on agent tool-call risk, with one weak spot you need to know about before you route on it. Across 240 hand-labelled …
Read more →agentgateway Token Exchange for AI Agents
An AI agent that calls tools for a user needs that user’s authority, and the usual shortcuts are a pasted personal access token or a shared service …
Read more →MCP Cross App Access Explained: The ID-JAG Exchange on the Wire
MCP’s Enterprise-Managed Authorization lets the corporate IdP decide which employees can use which MCP servers, by making the client trade an IdP-signed …
Read more →Rate-Limiting AI Agent Egress to Already-Allowed APIs
An egress allowlist decides whether an agent can reach a destination. It has no opinion about how often. On October 7, 2026, the Wikimedia Foundation disclosed …
Read more →Do AI Agents Need SPIFFE or OAuth?
They answer different questions. SPIFFE proves which process is calling, without a provisioned secret. OAuth scopes what it may do and, when a human is in the …
Read more →Per-Request Minted JWTs for Agent Backends with agentgateway
I built a demo where an AI agent calls a backend API and holds no credential for it. No API key in its environment, no key file on disk, no token. agentgateway …
Read more →AI Agent Security on Kubernetes: Which Controls Actually Held
I spent the last month building small, runnable labs for the ways AI agents break the infrastructure they run on: credentials, egress, audit logs, MCP tool …
Read more →Building Workload Identity Federation for AI Agents on Kubernetes
The easy way to give an agent pod access to an external API is a long-lived key in a Kubernetes Secret. The harder, better shape is the one I used in SPIFFE for …
Read more →Your Agent Egress Proxy Never Saw the DNS Query
An OpenAI research agent got its HTTPS requests to a public chatbot blocked by a web proxy. So it found a different way to ask the chatbot questions: it queried …
Read more →What Stopped AI Agent Probes of Government Data Sites
On September 24, Australia’s prime minister told reporters that an OpenAI agent had broken into a government Medicare statistics portal, accessed …
Read more →Your Artifact Registry Is a Two-Way Channel for Agents
Egress control for AI agents usually gets framed as a blocklist problem. Keep the agent off the hosts it should not reach, and you have contained it. This …
Read more →MCP Tool Poisoning: A Name Allowlist Is Not Enough
I built a four-container demo that reproduces an MCP server rewriting its own tool definitions mid-session, and puts two controls in front of it to see which …
Read more →Stopping an AI Agent Without Losing the Forensic Record
Two things get asked for the moment agents start doing real work. Stop it now, and keep everything it did. Those usually fight. The natural place to log what an …
Read more →A WAF That Reads the Prompt: OWASP CRS for LLM and MCP
A conventional web application firewall reads a URL, some headers, and maybe a form body. For agent traffic that is the wrong layer. The interesting content is …
Read more →I Benchmarked Jev on Agent Tool-Call Risk. Calibration Held.
TypeSafe AI launched Jev on 2026-09-15 with the usual launch-day multipliers — 193.6x faster, 444.6x cheaper than frontier LLMs — plus “zero …
Read more →agentgateway CEL Gotchas: Fail-Open + 403 Fix
I was writing a CEL authorization policy for an LLM route in Solo Enterprise for agentgateway : restrict which models a caller may reach, and refuse callers …
Read more →Agent Identity on Kubernetes: The Complete Series
An AI agent that calls a model, a tool, or another agent has to prove who it is. Almost every production failure I have written up this month traces back to …
Read more →Securing AI Agents in Production: The Complete Series
Almost everything in this series started the same way: a control that everybody agrees is the right control, applied to an agent, quietly not doing the thing …
Read more →Shadow MCP Servers: Track, Find and Stop Them Across Teams
Shadow MCP servers are hard to track across teams because most run as local child processes of an IDE or agent over stdio, with no network presence, and their …
Read more →Your Agent's LLM Key Survives Every Kubernetes Secret Control
I gave an agent pod an LLM API key the normal way, as a Secret projected into its environment. Then I turned on every Kubernetes control that sounds like it …
Read more →Keeping MCP Server Instructions Out of Your Agent's System Prompt
How do you keep MCP server instructions out of the trusted system prompt? Isolate them, cap them, bind any cache that stores them to server and caller, and pin …
Read more →The agentgateway Token Service Now Runs Without a Cluster
Solo shipped agentgateway enterprise 2026.9.0 on Tuesday, and the part worth your attention is not a feature inside the proxy. It is where the token service now …
Read more →Agent Audit Logs: What METR and Redwood Found
Two weeks ago METR and Redwood Research published their review of ~1,300 agent transcripts from the OpenAI / Hugging Face incident. Buried in it is a finding …
Read more →kagent Audience-Bound Tokens (RFC 8707)
kagent v0.10.0 went GA on September 4 . Buried in a long release list is a pair of environment variables that decide whether an agent’s credential works …
Read more →Why Proxy-Variable Agent Sandboxes Fail, and the CONNECT-Time Fix
On September 4, 2026, researchers published ~18,000 posts that autonomous agents left on a 25-year-old German wiki over six weeks, using it as a message board …
Read more →MCP Agent Identity: What Shipped, What to Use
The Model Context Protocol roadmap published on August 22 names agent identity and enterprise-ready security as a priority workstream, with four specific …
Read more →Your AGENTS.md Is Not a Security Control
An agent with a policy file that says “you must not delete customer records”, running behind an approval layer that denylists the delete tool, …
Read more →Which Controls Would Have Stopped the July 2026 Agent Intrusion?
An autonomous agent escaped its evaluation sandbox on July 9, 2026, and about twelve and a half hours of wall-clock work later it held cluster-admin on multiple …
Read more →SPIFFE for AI Agents: End-to-End Identity Demo
SPIFFE gives each AI agent a short-lived, attested workload identity (an SVID) instead of a stored secret, issued at runtime by SPIRE, rotated automatically, …
Read more →Your AI Agent Should Not Hold the LLM API Key
I built a demo where an agent calls an LLM and has no LLM credential anywhere in its process. No API key in the environment, no key file on disk, nothing in …
Read more →