Although originally not part of the plan (there’s a plan?), I was able to attend RSAC 2025 and see a couple of talks and hit the expo floor. I have to be honest; it was nice to see something, anything, not directly pushing generative AI hype (*cough* Google Next *cough*). Now don’t get me wrong, Agentic AI was front and center in a plethora of ways: semi-autonomous SOC agents, threat intelligence, and all over the early investment worlds. However, good old network and device security was visible, physical security had a place, and discussions about (still) getting the basics right could be heard. So I wanted to share with you the things I walked away with walking the expo floor and tracking announcements.
Top Trends and Announcements:
AI-Powered Threat Detection and Response: We’ve all heard this story. But at the conference companies like Cisco and Splunk demonstrated AI-driven threat response built into their products, while others like SentinelOne and Dataminr are introducing AI agents for proactive threat intelligence and faster incident analysis.
Generative AI INSIDE security tooling is now table stakes, and often Agentic: Almost every major vendor, notably CrowdStrike, Abnormal AI, and Securonix, are launching or expanding their offerings of autonomous AI agents designed to automate various security tasks. Improving efficiency in areas like employee training, data analysis, threat investigation, and threat hunting were the primary use cases, although some more independent vendors did claim to have fully autonomous agents to do the work.
Non-human Identity and Identity Management is a growth area: There were a number of new-ish vendors all focused on non-human identities and managing them alongside traditional identities. Solutions aimed to provide a unified view of human and non-human identities, identify and remediate identity-based vulnerabilities across hybrid environments, and offer complete lifecycle support for passwordless security. There were conversations with some real questions about how agents in an agentic framework would find and collect data with these identities without having overly permissive authorization and just-in-time access. This is clearly a growth space and somewhere we will see consolidation in the future.
Securing AI and Generative AI: Securing AI models and the data used to train them is becoming a critical focus, with a number of vendors looking not just at the training phase but also active attacks and red-teaming. Maybe you saw the announcement of Protect AI being acquired by Palo Alto, HiddenLayer had a much stronger showing than at past conferences, and in the early investment space EncryptAI and Promptfoo showcased strong offerings. Some red-teaming platforms and engagement vendors were even talking about attacking the gen AI models underpinning applications, although that was definitively more hit-and-miss. There’s still more work to be done, specifically around bringing industry intelligence to the threat models, but give us here at Generative Security a few more months and we’ll show you what that should look like.
Attack Path Analysis: It was encouraging to see the increasing adoption of attack path analysis in vulnerability reporting. I was impressed by Horizon3.ai for example on how they could tell me which fixes closed the most attack paths. However, it is notable and somewhat concerning to observe JupiterOne, a company that seemed to pioneer this area from my perspective, having a smaller presence at the conference compared to previous years.
Generative AI Acquisitions and “Complete” Protection: There were numerous announcements surrounding generative AI acquisitions, and I think this signals an early consolidation effort. While it’s still early to predict the ultimate market leaders, these moves offer valuable insights into the components that will likely constitute a comprehensive generative AI protection suite. I heard predictions from one investment panel that real consolidation would occur in 2 years, which I would agree with.
RSA Conference 2025 I think represented the breadth of security very well this year, with a good chunk focused on generative AI (no surprise there), but still having representation of the more traditional security aspects where AI takes a back seat. The advancements in areas like attack path analysis, identity security, and threat detection, alongside the emerging strategies for securing AI itself, was evident, and the rapid pace, honestly, a little scary. But in this industry, you have to be comfortable with being uncomfortable. As always, if you want to talk about any of this, feel free to contact at us questions@generativesecurity.ai. And if you’re looking at your own generative AI applications and wondering how to tell if they’re putting your information at risk, reach out. I’m happy to share what we’re doing but also connect you with some of the vendors in the community who can help do the things we’re not.
About the author
Michael Wasielewski is the founder and lead of Generative Security. With 20+ years of experience in networking, security, cloud, and enterprise architecture Michael brings a unique perspective to new technologies. Working on generative AI security for the past 2 years, Michael connects the dots between the organizational, the technical, and the business impacts of generative AI security. Michael looks forward to spending more time golfing, swimming in the ocean, and skydiving… someday.