Skip to content
Sign in
Artwork for The DevSecOps Talks Podcast
The DevSecOps Talks Podcast · Monday · 29 min

#111 - AI Agents: When Helpful Becomes Harmful

An agent asked to build a service found an Argo CD repository and deployed it to a cluster. Helpful automation gets harder to trust when a request leaves room for actions nobody intended. Mattias Hemmingsson and Paulina Dubas discuss AI-assisted compliance checks, unexpected deployments, evolving guardrails, company AI bans and whether regulation improves transparency or prices smaller competitors out. What you will learn How Claude turned AWS database checks into reusable audit evidence How an agent found a deployment path through Argo CD Why vague instructions and guardrail edge cases cause unwanted actions Why approving a hosted assistant does not make it local Where the hosts see benefits and competitive risks in regulation Full article and show notes: https://devsecops.fm/episodes/111/ Discuss the episode on LinkedIn DevSecOps Talks on YouTube

0:00-29:56
Play in

transcript

No transcript — this publisher did not publish one.

show notes

An agent asked to build a service found an Argo CD repository and deployed it to a cluster. Helpful automation gets harder to trust when a request leaves room for actions nobody intended.

Mattias Hemmingsson and Paulina Dubas discuss AI-assisted compliance checks, unexpected deployments, evolving guardrails, company AI bans and whether regulation improves transparency or prices smaller competitors out.

What you will learn

  • How Claude turned AWS database checks into reusable audit evidence
  • How an agent found a deployment path through Argo CD
  • Why vague instructions and guardrail edge cases cause unwanted actions
  • Why approving a hosted assistant does not make it local
  • Where the hosts see benefits and competitive risks in regulation

Full article and show notes: https://devsecops.fm/episodes/111/

Discuss the episode on LinkedIn

DevSecOps Talks on YouTube

links3

more episodes

All episodes