This blog post summarizes a webinar discussing the efficacy of AI agents in compromising cloud environments, comparing their success rate against canaries as early warning systems. The research showed that AI reached administrative privileges in an average of 14 minutes but was mitigated by canaries in 95.9% of incidents. The discussion is available on YouTube, Spotify, and Apple Podcasts.