---
title: AI Agent Security Starts with Architecture | Rasa x Lakera Report
image: https://info.rasa.com/hubfs/Rasa-Lakera-Social-V02.1b.png
---

[![rasa-purple](https://info.rasa.com/hubfs/RASA%20Main%20Theme%20-%20Act3%20Child/rasa-purple.svg "rasa-purple")](https://rasa.com)

[![rasa-logo-white](https://info.rasa.com/hs-fs/hubfs/newsletter/in-the-loop/rasa-logo-white.png?width=85&height=40&name=rasa-logo-white.png "rasa-logo-white")](https://rasa.com)

[![rasa-logo-horizontal-pink-digital-rgb (1)-2](https://info.rasa.com/hs-fs/hubfs/rasa-logo-horizontal-pink-digital-rgb%20(1)-2.png?width=85&height=45&name=rasa-logo-horizontal-pink-digital-rgb%20(1)-2.png "rasa-logo-horizontal-pink-digital-rgb (1)-2")](https://rasa.com)

[Start Referring Now](https://rasa.com/connect-with-rasa/)

[Talk with Sales](https://rasa.com/connect-with-rasa/)

[Talk with Sales](https://rasa.com/connect-with-rasa/)

- [Why Rasa?](https://rasa.com/product/why-rasa/)
- [Resources](https://rasa.com/resources/)
- [Get an Enterprise Demo](https://info.rasa.com/enterprise-demo)

[Connect with an Expert](https://rasa.com/contact-sales/)

**RASA X** LAKERA

## **AI Agent Security Starts With Architecture**

Independent red team assessment reveals critical differences in LLM agent security

Lakera, an independent security company, tested two AI agent architectures under adversarial attacks. Discover how design choices influence prompt injection vulnerabilities, data leakage, and overall system security.

### Get the free report

[![Rasa-Lakera-LP-Image-01](https://info.rasa.com/hubfs/Rasa-Lakera-LP-Image-01.png "Rasa-Lakera-LP-Image-01")](https://info.rasa.com/hubfs/2025%20Case%20Studies/%5BRasa%5D%20Lakera%20Red%20Security%20Assessment.pdf?utm_campaign=24381965-1025_LAKERA_CS&utm_source=landing_page&hsLang=en)

[![Rasa-Lakera-LP-Image-02 (1)](https://info.rasa.com/hs-fs/hubfs/Rasa-Lakera-LP-Image-02%20(1).png?width=996&name=Rasa-Lakera-LP-Image-02%20(1).png "Rasa-Lakera-LP-Image-02 (1)")](https://info.rasa.com/hubfs/2025%20Case%20Studies/%5BRasa%5D%20Lakera%20Red%20Security%20Assessment.pdf?utm_campaign=24381965-1025_LAKERA_CS&utm_source=landing_page&hsLang=en)

## Rasa’s structured agents showed strong security:

![vault-1](https://info.rasa.com/hubfs/vault-1.svg)

Zero prompt extraction or data leakage

Zero prompt extraction or data leakage

![shield-check](https://info.rasa.com/hubfs/shield-check.svg)

No content safety violations

No content safety violations

![lock](https://info.rasa.com/hubfs/lock.svg)

Resistant to jailbreak attempts

Resistant to jailbreak attempts

---

## 100%

of data leakage and content safety attempts fully contained

## 7x

fewer security failures vs prompt-driven baseline

## 0

data leaks, content safety violations and infra attacks

## Prompt-driven agents demonstrated exposure:

![folder-open](https://info.rasa.com/hubfs/folder-open.svg)

System prompts extracted

System prompts extracted

![warning](https://info.rasa.com/hubfs/warning.svg)

Customer data disclosed

Customer data disclosed

![shield-slash](https://info.rasa.com/hubfs/shield-slash.svg)

Content safety bypassed

Content safety bypassed

[![Rasa-Lakera-LP-Image-03](https://info.rasa.com/hs-fs/hubfs/Rasa-Lakera-LP-Image-03.png?width=996&name=Rasa-Lakera-LP-Image-03.png "Rasa-Lakera-LP-Image-03")](https://info.rasa.com/hubfs/2025%20Case%20Studies/%5BRasa%5D%20Lakera%20Red%20Security%20Assessment.pdf?utm_campaign=24381965-1025_LAKERA_CS&utm_source=landing_page&hsLang=en)

[![Rasa-Lakera-Image-04-v3.1a](https://info.rasa.com/hs-fs/hubfs/Rasa-Lakera-Image-04-v3.1a.png?width=2000&height=819&name=Rasa-Lakera-Image-04-v3.1a.png "Rasa-Lakera-Image-04-v3.1a")](https://info.rasa.com/hubfs/2025%20Case%20Studies/%5BRasa%5D%20Lakera%20Red%20Security%20Assessment.pdf?utm_campaign=24381965-1025_LAKERA_CS&utm_source=landing_page&hsLang=en)

[![rasa-logo-square-pink-digital-rgb](https://info.rasa.com/hs-fs/hubfs/rasa-logo-square-pink-digital-rgb.png?width=40&height=47&name=rasa-logo-square-pink-digital-rgb.png "rasa-logo-square-pink-digital-rgb")](https://www.rasa.com)

### Product

- Overview
- CALM
- Chat
- Enterprise RAG
- NLU
- Voice
- Agentic AI
- Multilingual AI
- Orchestration
- MCP
- [Pricing](https://rasa.com/pricing)

### Rasa for

- Developers
- Conversational AI Teams
- IT Directors

### Use cases

- [Customer Experience](https://rasa.com/use-cases/customer-experience)
- [Customer Support](https://rasa.com/use-cases/customer-support)
- [Sales Enablement](https://rasa.com/use-cases/sales-enablement)
- [Operational Efficiency](https://rasa.com/use-cases/operational-efficiency)

### Industries

- Finance & Banking
- [Insurance](https://rasa.com/industries/insurance)
- [Telecom](https://rasa.com/industries/telecom)
- [Retail](https://rasa.com/industries/retail)
- [Healthcare](https://rasa.com/industries/healthcare)
- [Travel, Transport, & Hospitality](https://rasa.com/industries/travel-transport-hospitality)
- [Government & Public Sector](https://rasa.com/industries/government-public-sector)<https://rasa.com/connect-with-rasa/>

#### Join our newsletter 📬

Stay up to date with the latest news from Rasa and conversational AI!

<https://x.com/Rasa_HQ/> <https://www.linkedin.com/company/rasa./> <https://stackoverflow.com/questions/tagged/rasa-nlu+or+rasa+or+rasa-core/> <https://github.com/RasaHQ/> <https://www.youtube.com/channel/UCJ0V6493mLvqdiVwOKWBODQ/> <https://angel.co/company/rasahq/>

© 2026 Rasa. All rights reserved. [Imprint |](https://rasa.com/imprint)[Privacy Policy |](https://rasa.com/privacy-policy)Security