The cybersecurity industry is facing unprecedented challenges as AI models become more capable. On this episode of Tech Disruptors, SentinelOne co-founder and CEO Tomer Weingarten talks about deploying security for AI agents and the evolution of the security operations center (SOC) with LLMs. He joins Bloomberg Intelligence analyst Mandeep Singh to discuss the impact of Anthropic Mythos, M&amp;A in cybersecurity and the changing nature of attacks as tools become more sophisticated.

SentinelOne CEO on Security Changes, AI Agents

Tech Disruptors

# Full Transcript

**Podcast:** Tech Disruptors

## Participants

- **Mandeep Singh** (Bloomberg Intelligence)
- **Tomer Weingarten** (SentinelOne)
- **Tomer Weingarten** (SentinelOne)

## Transcript

Hello and welcome to the Bloomberg Tech Disruptors podcast. My name is Mandeep and with me today is Tomer Weingarten, the co-founder and CEO of SentinelOne. Tomer, welcome to the podcast. Thank you so much for having me, Mandeep. Great. So lots to talk about. I mean, it feels like the cybersecurity market always has a lot going on

and maybe a good place to start off would be really how you see this whole kind of Mythos impact from Entropic in terms of how you as a pure play cybersecurity vendor has been looking at it from a disruption standpoint

and also, you know, the impact that you see from a model like that for the security industry as a whole. And let's start by saying that, you know, there are quite a few models right now that are showing pretty significant capabilities along the lines of, I think, what we, you know, what we've seen from mythos according, you know, to third-party research.

I think, you know, the UK AI Institute, AI Security Institute actually tested both GPT 5.5 and mythos in the same benchmarks and they perform pretty similarly, I would say. And I think if I take a step back, you know, Mythos, I think, is a nice reminder to all of us that these models are going to be better and better and better with time.

And I think that what you're seeing is this, what I would define as the continuous exponential improvement of these models. And I think that that's the key thing for all of us to understand and specifically in cybersecurity. These models are getting more and more capable in cybersecurity and outside of cybersecurity.

I think it's a derivative of their ability to complete tasks in a longer horizon, in a more predictable manner that also makes them, I think, very capable in, you know, conducting cyber attacks as well. But at the same time, I think focusing specifically and only on the vulnerabilities and the discovery of vulnerabilities

and also doing a disservice for cybersecurity. There's been a lot of talk about, you know, how these models are helpful in finding vulnerabilities in source code. But that's not the only issue we have in cybersecurity. And, you know, throughout many, many years, exploitation has been only one vector of attack. And there's plenty other ways, you know, for states, you know, state-sponsored actors

or just cyber criminals to compromise devices, networks, and people. And I would also say that the one thing that I don't hear enough talk around is the growing attack surface with AI agents themselves. Because you don't necessarily need a zero-day exploit to actually convince or change or inject instructions

into a running autonomous AI workload. Now, we're all rushing to deploy more and more AI agents and, you know, more AI workloads. We're very focused on vulnerabilities. But are we focused e...

SentinelOne CEO on Security Changes, AI Agents

Transcript

Built for investment professionals

Every day you wait, your competitors are finding
edge in conversations you're not hearing.

SentinelOne CEO on Security Changes, AI Agents

Transcript

Built for investment professionals

Every day you wait, your competitors are findingedge in conversations you're not hearing.

Every day you wait, your competitors are finding
edge in conversations you're not hearing.