Agentsec Weekly: A Newsletter on AI Agent Security
After a few quiet months, I’ve started Agentsec Weekly, a free weekly newsletter on AI agent security. alphasec isn’t going anywhere.
It's been quiet for a few months here. Most of what I've been reading, testing and building lately has been about AI agents: how they authenticate, what they're allowed to do, and how they get compromised. That material moves faster than I can turn it into long-form posts, so I've started a separate weekly newsletter for it.
What Agentsec Weekly Covers
Agentsec Weekly is a short weekly briefing on AI agent security. Each issue covers:
- Vulnerabilities and incidents in agents, MCP servers and coding tools
- Standards work on agent identity, delegation, and related protocols
- Security features shipped by the major cloud and developer platforms
- Open-source tools worth trying, and podcasts worth listening to
Here is the first issue, if you prefer to get straight to it.
alphasec
How It's Made
AI does the first pass: it checks a fixed list of primary sources (advisory databases, standards working groups, vendor changelogs and research blogs, and podcast transcripts), then drafts summaries. I review every issue, check claims against the original sources, and decide what makes the cut. Every item links to its source, and ideas from podcasts are credited to the people who said them.
alphasec Isn't Going Anywhere
The newsletter doesn't replace this blog. Agentsec Weekly is the quick weekly signal; alphasec retains the long-form work: personal experiences, best-practice guides, and hands-on write-ups where I've broken something and figured out why. Some newsletter stories will probably turn into posts here. So please stay subscribed here.
If agent security touches your work, subscribe to Agentsec Weekly too. It's free, independent, and lands every Sunday.