Articles on AI agent security, adversarial testing, posture management, and enterprise AI risk.
Two lines of Python attach the Humanbound Firewall to a LangChain agent. This setup guide walks through the policy file, the Tier 3 judge model, and rolling enforcement out in log mode before you switch to block, using a customer support bot as the running example.
Hugging Face dropped Opus when its guardrails blocked incident work. hb 2.13 lets you red-team with any OpenAI-compatible model. Five OpenRouter setups compared on empty replies, cost and speed, and why you should pin the model and the host.
We gave nine models a pricing agent's job and a competitor page one link away from a trap. Four leaked a confidential unit cost on every run. Here is what 1,350 runs say about how much a system-prompt guardrail actually buys you.
A one-sentence prompt change passed every normal check. A red-team gate in GitHub Actions caught it for 26 cents. The full workflow, real numbers from three runs, how to cap spend, and what to do when the build goes red.
Every AI agent your teams shipped can act on its own, including the ones nobody approved. Here's how CISOs prove agent security before a regulator or the board asks.
Humanbound's official GitHub Action gates a build on adversarial AI agent security tests, failing the workflow past a severity threshold and uploading findings to GitHub's Security tab as SARIF.
A working LangChain agent isn't a tested one. Wrap it in FastAPI, red-team it with Humanbound, and find out what it actually does under attack.
Ayan built a customer support agent in an afternoon, then pointed Humanbound at it and watched it fabricate a settled $185 refund for an order that does not exist. A hands-on walkthrough of running Humanbound's local adversarial testing against your own agent.
Splunk is where most SOCs already live. Here's how Humanbound streams AI agent security findings into Splunk as structured, HMAC-signed webhooks, so agent security gets the same real-time alerting and incident response as everything else the SOC already handles.
PyRIT is Microsoft's open-source, research-grade red teaming framework. Here's how a point-in-time PyRIT engagement and Humanbound's continuous, compliance-mapped monitoring plug into each other, including how PyRIT's findings can train the Humanbound Firewall directly.
Promptfoo and Humanbound aren't competing for the same slot. Here's how Promptfoo's CI red teaming and Humanbound's continuous, compliance-mapped monitoring plug into each other, including how Promptfoo's scan results can train the Humanbound Firewall directly.
In 2017 I got a CVE for an unencrypted smart bulb. Nine years later, AI agents are shipping with the same gap: nobody tried to break them before launch. Here's how to break your own agent this afternoon, before someone else does.
A community library of agent attack scenarios, each mapped to the OWASP Top 10 for Agentic Applications and paired with the guardrail that closes it. Open, credited, and free to contribute to.
We released humanbound-firewall under Apache-2.0. A multi-tier runtime defense for AI agents, with each layer inspectable, escalating on uncertainty, and trainable on your own adversarial test data.
AI security maps to two different markets: AI for security (AI4Sec) and security for AI (Sec4AI). The Claude Mythos Preview made the distinction unmissable. Here is how to tell which one you are actually buying.
Moderation APIs catch harm and injection attempts but fail to enforce domain-specific policy. A cross-domain evaluation shows why production LLM systems need both moderation and policy reasoning layers.
The security industry is applying shift-left to AI agents. But AI agents aren't deterministic. The gap between testing on deploy and staying secure in production is where risk accumulates.
The AI agent security market is fragmenting into enforcement, identity, and control plane vendors. But the incident data tells a different story: most agents ship without any adversarial testing at all.
Agent security incidents keep sharing one root cause: agents deployed without adversarial testing. Enforcement is needed, but it is phase 3 of a lifecycle most organizations are entering at phase 1.
AI agent security won't be adopted through better dashboards. It'll be adopted when it disappears into existing workflows. Embed adversarial testing into your terminal via MCP - security becomes a conversation, not a context switch.
Why point-in-time audits fail, and what a mature AI security programme actually looks like. A continuous assurance framework addressing visibility, testing, and operations gaps in enterprise GenAI deployments.
Claude Code Security didn’t kill security tooling—but it did signal that frontier AI labs are now security providers, and that our current models don’t cover the behavioral, contextual, and systemic risks of autonomous agents. This is where security needs to go next.
Your CISO dashboard tracks vulnerabilities across infrastructure, apps, and cloud—but may miss the fastest-growing risk surface: Shadow AI. See how it shows up, why legacy tools miss it, and how Humanbound reveals AI use in a governed, continuously monitored registry.
