We told our agent to hack another agent. It found a way in. Then another. Then another. And we can probably hack your agent too. Today, we’re launching Respan Security.

1/ It's way too easy to exploit AI agents. We spent two years building observability for them. The more we saw, the harder that truth became to ignore.

2/ Agents now have access to sensitive data, tools, and infrastructure. Security can’t stop at the application layer. Teams also need to test what users can convince the agent to do.

3/ Respan tests through the same interface available to users. It doesn’t need source code or internal access. It behaves like an adversarial user and checks whether it can make the agent break a security rule.

4/ Every assessment starts with reconnaissance: tools, refusals, guardrails, prompt leakage, and likely attack paths. What Respan discovers determines which attacks run.

5/ Each probe runs in a loop: Attack → response → judge outcome → update campaign → choose next attack The result informs what Respan tries next. Respan doesn’t just give up after one refusal.

6/ The final report shows which attacks worked, the evidence behind them, their severity, and what to fix. Hack your agents before attackers do. Run a free test today:

It’s too easy to exploit agents nowadays

shipping an agent without red teaming it is basically deploying an API that only honest users are allowed to call

Agent security is going to become a much bigger problem than most people realize Excited to see where this goes 👀

security should be the top topic today

So excited to see more focus on AI agent security, doing great work here!

Security is always important.

Way to go!

🚀 sick

Agent-to-agent attacks are the security blind spot most teams aren't prepared for yet. Red-teaming your AI before someone else does isn't paranoia — it's the new baseline.

nice

😍

We

Cool!

So we're at 'my agent can hack your agent' already? Wild. I laughed, then checked my stack. This layer was overdue.

Good

un-sandboxed code execution makes coding agents trivial to hijack

nice

thanks for sharing.

nice

Very interesting. I'd love to see a real-world case study showing how an agent was compromised and how Respan Security detected or prevented it.

Easy😎

g

hi

This is impressive

Congrats @RespanAI — this ranked #3 on Launch Archive today.

This is an important direction. We spend a lot of time building capable agents, but not nearly enough time thinking about how they'll be attacked. Great to see security getting the attention it deserves.

Great!

A great reminder that agent security can't be an afterthought. As autonomous systems become more capable, continuous adversarial testing is essential to build resilient, trustworthy AI. 🔒🤖

hgjfj

Wow, what a scary world we live in

so impressive.

great