Cloud Resilience Becomes the Real Cybersecurity Strategy
Every organization should assume it will be breached. That reality makes cloud resilience the security strategy that matters most. AI agents move fast, adversaries use them too and mistakes are inevitable.
Firefly Co-Founder and CEO Ido Neeman examines that shift with Alan Shimel at KubeCon + CloudNativeCon in Salt Lake City.
Regulators Want Proof of Recovery
Neeman points to Europe’s Digital Operational Resilience Act, or DORA. It asks organizations tied to European financial institutions to prove they can recover. Thinking about an outage is no longer enough.
Regulators want to see a tested plan. That plan must cover hackers, cloud outages and accidental deletions alike.
Good Agents Make Mistakes Too
Not every incident comes from an attacker. Neeman shares a story about an AI assistant that sent him to a closed restaurant. That mistake left him hungry.
The same kind of error in a cloud deployment could take an application down. Thousands of customers would feel it. Understanding an entire cloud or Kubernetes environment is far harder than checking restaurant hours.
He argues every agent should be tied to a human owner. Someone must stay accountable when things go wrong.
Mapping the Cloud for Fast Recovery
Firefly builds a context lake of a company’s full cloud estate. It spans cloud providers, Kubernetes and vSphere, plus tools like GitHub, GitLab and Okta. It captures how resources are built, configured and connected.
From that map, Firefly identifies the infrastructure behind each application and codifies it. Teams can roll back to an earlier state or redeploy to a new region. They can also move to a clean account after a breach.
Neeman says AI lets Firefly rebuild complex environments in minutes rather than days or weeks.
Resilience Posture Management
Shimel notes that resilience often pairs with containment. A smaller blast radius means less to restore.
Neeman adds that teams now need resilience posture management. They must know where they are weak and harden those parts first. For enterprises, cloud resilience has become a board-level concern.
Transcript
Hey everyone, we're back here live at KubeCon in Salt Lake City. It's been a good couple of days here. We've spoken to so many people.
Great conversations, so much going on. I got another one I want to talk to you with. If you've watched our CUBE coverage in the past, you know my friend, Ido Neeman.
Ido is with Firefly. First of all, let's welcome him back. Ido, it's great to have you here, man.
Thank you for having me back. All's good? Yeah.
Very good. But, not everyone watches everything I do. As much as it upsets me.
Most of them do. Most of them do. Most of them.
But a small minority. But for those who maybe don't know you and Firefly, well, let's start with you. Give them your story.
My name is Ido Neeman. I'm the co-founder and CEO at Firefly. Firefly is the cloud resilience platform.
We can get to it later. I live in Israel, and we operate teams across US, Europe, Middle East, and Israel, of course. Of course.
Yeah. And what does Firefly do? Firefly is the cloud resilience platform.
Basically, we help teams and companies understand the cloud, control it, and recover it from incidents in minutes, not in weeks. Love it. Before we go any further, for people who want to get more information on Firefly, it's firefly dot...
It's firefly dot, you guessed it. AI. Exactly.
ai. I didn't see that ending. com is the best dot-com.
Say it again. Say it loud for the people in the back. com.
com But you know that these days, if you don't have the dot AI- AI ... something is broken. Well, I do have here somewhere.
ai. ai. Yeah.
I got Textrong IT. We try them all. Yeah.
But you mentioned resilience. All kidding aside, resilience is a term, it was high on everyone's radar. And now with AI, quite frankly, everyone is so consumed with agentic and AI, we assume resilience.
And you know what happens when you assume, right? You make an a*s of you and me. And so shouldn't assume resilience.
Talk to me about resilience. Let's build on this everyone thinks agentic, okay? Everyone think agentic, but also everyone is becoming more aware of the risks that are associated with- That's true ...
agentic capabilities, right? We saw what happened with PHP two weeks ago. We see what happened with other breaches.
Everyone must assume they will be breached, they will be compromised. So if you understand that the agents and AI capabilities are moving fast, and adversaries are leveraging them for their intentions, which are malicious. And even if you're not malicious, if you just test a model in a sandbox and it breaks out, and it goes hacking website, what happens if very bad guys train agents to do it, right?
Yeah. Everyone will be breached. In 2026, 2027, the only true cybersecurity strategy is the resilience strategy.
Yeah. " Get back up, yes. Right?
This is our reality now. I never realized that was Confucius, but okay. Yeah.
But it is where we are now. I have a similar... Tell people all the time, we're not going to patch our way out of this, right?
Yeah. You're not. So the only answer then is what happens when?
What do I do when it happens? Not if, when it happens, what do I do? What's my resiliency?
So I think the nice thing about it is regulation. I'm supportive of the free markets, and not only about regulation, but I think regulators are telling us the way to do. Mm-hmm.
Regulators just introduced in Europe a framework called DORA, the Digital Operational Resilience Act. And it calls upon everyone that has any relationship with financial institutions in Europe to prove they can recover. To prove that when a hacker hacks your account, an agent accidentally deletes your stuff, a cloud outage happens, like we saw two or three weeks ago, Google Cloud had a major outage in London.
Suddenly, Revolut cards are not working. HSBC is down, right? Those are great organizations, but they won't plan for this type of scenario.
" Not show us that you've thought about it. Show us your plan. Your plan.
You test it, you've proven resiliency- Yeah ... and now you go back to your business. In AI era, everyone, we are moving so fast, and attacks are becoming so frequent and aggressive that one must plan for it.
But you know what? I started with talking too much about the adversary malicious hacker. I told you, I think, just before we started, that I came here from McDonald's.
No. Why is it? I love McDonald's, but once I'm in Salt Lake City, I want to try something else.
I don't see a lot of McDonald's here, I'll be honest with you. Yeah, not a lot of them. They got other ones.
Yeah. But go ahead. I digress.
So what happened? I asked my Instinct agent to recommend a good restaurant. It sends me there.
Now I get there, it's closed. And I'm asking Instinct, "Hey, is it closed now? " "Oh, I made a mistake.
" I'm hungry. I'm in a problem. But if I would do the same using an agent, code to deploy to my cloud, and it made a mistake, now my application is down, and I have thousands of customers angry, and I need to rebate them, and I'm in a big problem.
So agents are moving fast. Agents are everywhere, but we all know it, agents make mistake. Yes.
Finding if a restaurant is open is dramatically more simple than understanding an entire cloud or Kubernetes environment and change it and provision it and control it. So agentic cloud ops will cause lots and lots of good things alongside some problems and outages, and you must be ready for this event. So it's not just the bad guys.
Those are the good agents that are alongside us. No, there are no bad agents. You know what I mean?
These things are going to happen. It's part of the process. But I want to talk about, okay, so how do you provide resilience?
Firefly maintains a context lake of your entire cloud. We connect everything that you run. AWS- Is it like a digital twin?
It's a different concept, okay? But think about it, we have this context lake. All your cloud estate, mentioned it could go to Oracle and Nebulas and Kubernetes and vSphere, but also your Akamai and your GitHub and GitLab and Okta.
Everything is part of your platform. We know everything about it. How everything is built, configured, connected, what are the relationships, what are the dependencies?
And then we take this context lake, and we map applications out of it. We know all the infrastructure needed for this application. It will bake it into our technology, which uses some sort of infrastructure as code.
Now we have it, it's modularized, we can move it around. We can move it to the same account for 17 minutes ago, from 17 days ago, 17 weeks ago. Or we can redeploy to a new region if there's a cloud outage.
If you get hacked and you need a clean account or a clean room or an IRA, we move you there. So it doesn't matter what your use case, if you need to recover in minutes instead of weeks- Whether it's ransomware, an agent, any of it. Exactly.
Or, we still have people making mistakes. If I make a mistake- Yeah. and I deleted my, let's even call it my Entra ID's tenant.
But tell the truth, you wouldn't blame it on the agent. Of course I will. Yeah.
But this is why every agent must be associated with a human owner, right? Because at the end, as management, sometimes we need to talk to a person that is responsible and get them solve a problem for us. I love it.
Now, so you've adapted to this whole agentic thing going on here. But the other issue with agentic is scale, speed, and breadth, right? Scale.
What have you had to do to scale Firefly to this new world with how fast everything goes now? It's just the scale is off the chart. So if I'm honest, I don't think we had any issues with scaling Firefly.
Firefly was built to manage and control- Very large and massive multi-cloud estates, but the other way around. Now with AI, we can recover an estate that would have taken us two days in just 14 minutes. Yeah.
Because let's say you have a complex environment. Sometimes you can have 1,500 different API addresses that all talk to each other, and have the certificates, and have the ACLs, and need to correspond with IAM roles and IAM policies. How do you manage it all?
It's not human scale. Only AI scale do it. So if you have proper secure agents running with the right guardrails, the right policies, and with resiliency, then you can rebuild their environments in minutes instead of weeks, because we see it.
AI agent mistakes and hackers take down companies, and it takes them weeks to recover. Why? Because it's very complex.
If you have Firefly and the power of AI harnessed to your benefit, you can do it in minutes. I love it. Firefly AI.
Well, let's talk here about KubeCon a little. Any announcements here at the show? Or what are you seeing?
What are you hearing? I'll be honest, I was pretty heads down focusing on Firefly, our customers, our partner. I think that resiliency is top of mind for many right now.
I can tell you that across our customer base, especially when we talk to the enterprise level customers, we see resilience become a bold level visibility issue. Absolutely. I'm going to tell you something else.
Where I see resilience is usually partnered with containment. So I've got to contain an attack. I've got to contain malware.
I've got to contain my blast radius. And that plays into the Firefly, because the less I need to restore, the easier it is for you to restore it- Exactly ... versus having to restore the whole thing.
Exactly. And also, it goes back to simple cloud concept of posture management. Everyone talks about security posture management and cloud security application.
Segmentation, micro-segmentation, all of these things. But it's also now everyone must understand the resiliency posture management. Where am I vulnerable?
Because everyone has vulnerability. You need to understand it. You need to understand what happens if this small part falls apart for you, and then build strength and harden this specific part for you.
This is one of the things that Firefly helps customers understand. I love it. We're about out of time.
ai, resiliency. Remember that. When you think resiliency, think of Firefly.
Hey, we're coming back. We've got a lot more today. It's busy here.
We'll talk to you later.