Ahead of RSAC 2025, CalypsoAI CTO James ‘Jimmy’ White talks all things Agentic with communications lead Gavin Daly. They cover Jimmy’s insights for this year’s conference, what it means to be a top 10 finalist in the Innovation Sandbox contest, and more.

What Are You Expecting to See at RSAC?

Last year’s RSAC was all AI. There is this famous James Joyce quote about the puzzle of crossing Dublin without passing a pub, which is an impossible mission. That was the game last year – crossing the Moscone Center without passing a sign that said ‘AI’. 

This year, I think, will all be Agentic – the manifestation of AI into both virtual and kinetic impact. So the ‘action’ part of AI will be Agentic. Of course, as Agentic gets into more places – both virtual and physical – how do you protect that? How do you secure that? And so, with RSAC being the biggest security conference, that will be the biggest talking point. 

So What Are The Priorities as Agents Go From Hype To Reality?

To keep it really simple, an agent is made up of three things: a purpose, brain and tools. The purpose is the job you’ve given it; the brain is an AI model or models; and the tools could be digital or physical. So what it can actually do may be virtual or physical, kinetic or not.

There are two places you need to protect, two attack surfaces there. The first is the thought phase: if the agent starts thinking bad things, or thinking things that we don’t want it to think, we need to stop it in its tracks. And if you do that, you will rule out a lot of unintended consequences. 

The second thing is the action phase: every now and again, the agent will do something that is a good thought, but the action is wrong. So when the brain tells the tool to do something, you also have to police that to make sure that action isn’t really bad. Because the idea was good, but the implementation was bad. 

They are the two places you need to provide protection: at thought and at action. That’s an incredibly complex, fast-moving area to try and protect, and so the solution has to be as unique as the problem.

How Would You Rate Awareness of the Risks?

Organizations want to move as fast as possible in the most efficient way, so I think the intent is good. How you do that, that’s where the devil is in the detail.

For example, if agents are autonomously providing information at scale and at speed to humans to make a decision, that’s somewhere where you can be less prescriptive on the security requirements. But if agents are actually having autonomy to make actions, that’s a whole different ballgame, right? 

The big question I have around all Agentic use cases is around the identity of the agent: Is the agent a proxy? Is it an actor? Is it an information gatherer? Even read-only use cases can cause problems: an agent reading a select statement on a database table that’s 100 rows long is fine; one against a billion rows long can cause a DB lockup, right? So there’s certainly a balance to be struck.

What Will People Be Hearing From CalypsoAI at RSAC?

Effectively, two key messages. The first is that we protect Agentic as well as regular AI; protecting Agentic at thought and at action pace is our defensive capability. 

And second, our big product launch is our Inference Red-Team solution, and that’s Agentic-driven. So, as well as releasing 10,000 new attacks a month, we actually have a team of our own agents that conduct red team assessments on your AI systems and give you a lovely report that shows you all of the flaws. For the record, we can break every single model right now.

For us, it’s about exposing weakness in AI systems with our Red-Team and highlighting that before you go to production, and then filling in those gaps with our defenses, with custom controls. Every good sports team, if you’ve got good offense, good defense, you have a pretty good chance of winning the game. Of course, it’s a continuous process because you don’t know what new attacks are going to launch the next day; that’s nothing new to infosec professionals.

What’s most important for me as CTO is that our product is real. There’s a lot of hype at RSAC, a lot of Venus flytrap stuff where people say ‘we’re launching this product’ but information is lagging. A lot of big companies try to use this as an indicator of what to build and invest in. Whereas when we’re there, we’ve got a real product. Everything we say is true. It’s there today.

Any Tips for Maximizing Your Time at the Conference?

• First, when you go to a booth and you hear the prepared message, politely interrupt that message and ask: ‘can I speak to a technical representative of your team?’ I always think of Hollywood events where the reporters want to get past the agent and talk directly to Tom Cruise. So you want to skip through to the technical person.

• Tip two is, when you have access to a technical person, ask them three specific questions. If it’s Agentic, for example, did you role it yourself or are you using a framework? And if so, which framework? If there is not an answer available for that, Agentic is not in the system – no matter what the marketing says, it’s not in the system. 

• The third tip is to talk about the roles. So we talked about those three things that an agent is: it’s a purpose, it’s a brain, it’s a tool. If any of those three things aren’t present, it’s not an agent. So ask those three questions: in your agent, what’s the purpose? What brain are you using? And what tools are available? They may need to be respectfully cautious on answering because it could represent IP, but you can hopefully spot BS there or not. If they’re willing to share abstract answers for those three questions, I think that’s fair enough.

There’s one thing I always say I’ll do and I forget: write down your top five companies you spoke to and record some details. Because I guarantee you, on your way home, you’ll be thinking to yourself: What was that company? What do they offer again? If there’s other people from your company or your network, compare notes. You’ll have missed a bunch of vendors and they’ll have missed a bunch of vendors. And truing up those notes and sharing is a really good way to have an action list post-RSAC.

Finally, Tell Us About Being the RSAC Innovation Sandbox

We’re a Top 10 Innovation Sandbox company, which is amazing. It makes this year very different for us. The folks who run the competition are unbelievable, they’ve got everything mapped out: where we’ve got to be, our timeline, what we need to do to prepare to present on the Sandbox stage. 

We’ve got an extra demo booth because we’re Top 10 finalists so we’ve got to have people at our actual booth on the RSAC floor and at our Sandbox booth. We’ll have a big CalypsoAI team this year. We’re going to get there early, be super-prepared and do a great job at the competition; then have ‘phase two’ of it be regular RSAC, where we’re at the booth meeting customers and prospects. It’s always super exciting for the whole company and more exciting than ever this year. 

Watch the video to get Jimmy’s full take on RSAC 2025, the rise of Agentic AI, and how CalypsoAI is leading the charge in real-time, inference-layer security. If you’re heading to RSAC, come see us at Booth #967, where we’ll be showcasing our latest advances in Agentic AI security.