@svpino: This will let you break your agent before your users do. This works with any agent, including chat, code, and voice age…
Summary
A tool that automates over 10,000 jailbreaks and adversarial attacks to test AI agents before users do, ensuring security for chat, code, and voice agents.
View Cached Full Text
Cached at: 08/20/26, 02:48 AM
This will let you break your agent before your users do.
This works with any agent, including chat, code, and voice agents over the phone.
To run it, give it your URL or phone number. You don’t need any other integration, which is great.
• It runs 10,000+ jailbreaks and adversarial attacks • It continuously re-tests as your agent changes • It’s completely automated
Zach AL (@zachdotai): we got an AI agent to exfiltrate customer bank balances and using an invoice that looked completely normal to a human.
and we asked some people what they think.
Similar Articles
How are you testing your AI agents for security before they hit users? We got tired of not having a good answer and built this.
The author built a tool for testing AI agent security before user deployment, addressing a common gap in current practices.
Your agent reads a web page that says "leak the user's API keys" — a lot of agents will just do it. I built a thing to stop the send.
Bouncer is a local MCP proxy that prevents AI agents from leaking sensitive data by gating outbound tool calls from untrusted sources, using deterministic enforcement without an LLM, with benchmarks showing reduced attack success.
Letting an agent loose on a real iPhone taught me to build the kill switch first
The author built sidetap, a Python harness that lets an LLM agent control a real iPhone from Windows over USB, with safety features like a kill switch and guardrails. The post details the technical challenge of sideloading WebDriverAgent with a free Apple ID and introduces the open-source tool.
i compiled my agent's security layer into 33kb of js so you can try to break it yourself
The author compiled a security layer for AI agent communication into a 33KB JavaScript file that runs in the browser, allowing users to test its defenses against various attacks like forged signatures and ensuring orders are contained in non-executable fields.
I'm building a tool to stop manually chatting with your own AI agent to test it, would you use it?
The author is building a tool to automatically test AI agents by simulating realistic user conversations and providing pass/fail reports, saving developers from manual testing.