# @awseventschannel on YouTube

- **Type:** Video
- **Original URL:** https://youtube.com/watch?v=cw26_XNmnJw
- **Gondola URL:** https://gondola.cc/posts/63513932-awseventschannel-youtube
- **Thumbnail:** https://img.gondola.cc/tr:w-,h-,fo-auto/postThumbnails/fbbddbdfa6.jpg
- **Posted:** 2026-03-27T05:37:47.000+00:00
- **Account Owner:** AWS Events (@awseventschannel) — https://gondola.cc/awseventschannel

## Caption

🛡️ What if unauthorized data access was architecturally impossible—not just discouraged? 

AI agents that touch production databases create a new kind of security surface. Prompt-level instructions like "don't access other users' records" might work in a demo, but they're not #Cybersecurity controls—they're suggestions that a well-crafted injection can blow right past. 

This stream builds the alternative from scratch: 

🔹 See what a prompt injection attack actually looks like against a database-connected #agenticAI agent (it's fast & it's ugly) 
🔹 Wire up identity propagation so #AmazonDynamoDB & Amazon Aurora enforce per-user row-level isolation—no matter what the agent tries 
🔹 Set up Amazon CloudWatch monitoring to flag anomalous access patterns in real time 

Stop hoping the prompt holds. Start making the wrong query return zero rows.

## Stats

- **Views:** 1,508
- **Likes:** 34
- **Shares:** 0
- **Comments:** 5

## Credits

No public credits are recorded here. This does not identify who made the post.

## Tags

cybersecurity, agenticai, amazondynamodb

---
Copyright (c) Gondola