10 Hot AI Security Startups To Know In 2025

Here are 10 startups that are on our radar for bringing new ways to secure GenAI-powered applications and AI agents.

AI Security Startups To Know

The rapid advancement of AI and agentic technologies has been dominant in the cybersecurity space in 2025 — with early-stage startups playing a central role in helping to protect the emerging AI capabilities. While numerous AI security startups have raised funding and launched their products commercially, top security vendors have also snapped up many of the newer players in a major wave of startup M&A.

Those acquisitions have, in turn, led to even faster acceleration in the segment. Protect AI, for instance, brought aboard Channel Chief Nick Crouch in March — and just weeks later the AI security startup was acquired by Palo Alto Networks for $700 million. Now, Crouch and other executives from Protect AI, including CRO Ralph Pisani and Head of Marketing Alexandra Bush, have joined their second AI security startup of the year by recently landing similar roles at Noma Security. Without a doubt, “the AI security market is exploding,” Bush told CRN. And with the massive need for up-to-moment innovation for securing the fast-changing AI landscape, “this space isn't going away,” Bush said.

Notably, early-stage AI security vendors such as Noma are also increasingly looking to work with the channel, with the startup expecting to soon launch a partner program in the U.S. after raising $100 million in July, Crouch said. There’s no question that, for a startup in a rapidly evolving space like AI security, “the fastest route to market is channel,” he said.

As part of CRN’s Cybersecurity Week 2025, we’re highlighting 10 AI security startups that have been on our radar for recently unveiling new ways to secure GenAI-powered applications, AI agents and LLMs.

What follows are the details on 10 hot AI security startups to know in 2025.

Astrix Security

Founded: 2021

CEO: Alon Jackson

In September, Astrix Security unveiled its new offering focused on enabling “secure-by-design” AI agents. The AI Agent Control Plane platform provides agents with credentials that are short-lived and “precisely scoped,” the company said in a news release. Key capabilities include just-in-time access, leveraging the principles of least privilege, as well as the ability for developers to deploy AI agents from their usual tools and centralized control after deployment, according to Astrix.

Aurascape

Founded: 2024

CEO: Moinul Khan

In April, Aurascape emerged from stealth with its offering that secures user activity across “thousands of AI applications and tools,” with enhanced visibility and strict controls for “every AI interaction,” the company said in a news release. Core capabilities include risk analysis, decoding for prompts and responses, data protection and automated remediation. Aurascape also announced in April that it had raised $50 million in funding from Mayfield Fund and Menlo Ventures.

Descope

Founded: 2022

CEO: Slavik Markovich

Descope is seeking to become a go-to agentic identity provider for the era of AI agents, according to Descope co-founder Rishi Bhargava. In August, the company debuted its Agentic Identity Control Plane, with capabilities that can enable the deployment of policy-based governance, identity management and auditing for AI agents and Model Context Protocol systems, the company said. Descope also announced in September that it had added $35 million in new funding, closing its seed round at a total of $88 million.

Irregular

Founded: 2023

CEO: Dan Lahav

In September, Irregular disclosed details about how it’s providing security for highly advanced AI, through what it calls the industry’s “first frontier security lab.” The startup said that its research platform is capable of running controlled simulations on advanced “frontier” AI models, through testing for misuse in security operations as well as assessing their resilience against cyberattacks. Irregular also announced it September that it has raised $80 million in funding led by Sequoia Capital and Redpoint Ventures, with other investors including Wiz Co-founder and CEO Assaf Rappaport.

Noma Security

Founded: 2023

CEO: Niv Braun

Noma Security offers a platform for controlling AI and agentic risk with a major focus on providing continuous discovery for AI assets and agents. The platform provides a full AI inventory through automatically discovering where AI applications and agents are being built, as well as determining which data and systems they can access, according to Nick Crouch, senior director of channel and alliances at Noma. Other key capabilities include AI security posture management and risk prioritization as well as AI runtime protection, the company said. In July, Noma Security raised $100 million in a Series B round of funding led by Evolution Equity Partners.

Operant AI

Founded: 2021

CEO: Vrajesh Bhavsar

In June, Operant AI unveiled an expansion to its runtime AI defense platform, AI Gatekeeper, with the debut of its MCP Gateway. The offering provides “comprehensive” protection for MCP (Model Context Protocol) applications, with capabilities including real-time discovery and detection, the company said. The MCP Gateway offering ultimately provides “defense for every layer of the MCP stack,” Operant AI said in a news release.

Promptfoo

Founded: 2024

CEO: Ian Webster

Promptfoo offers an open‑source framework that can be used to enable AI red‑teaming and security testing, with a focus on helping security teams and developers to protect their LLMs and GenAI-powered applications. Key capabilities include detection and mitigation for prompt injection attacks and data leakage, as well as for insecure tool use, the company said. In July, Promptfoo announced raising an $18.4 million Series A round of funding led by Insight Partners.

Relyance AI

Founded: 2020

CEO: Abhi Sharma

In September, Relyance AI announced a major new feature for its AI-native data security platform, with the introduction of what the company calls the industry’s “first data defense engineer.” The capability provides continuous assessment of AI interactions to “understand, monitor, and protect thousands of data journeys” — ultimately providing autonomous enforcement of policies in real time, the company said in a news release.

Straiker

Founded: 2024

CEO: Ankur Shah

In August, Straiker expanded the functionality of its Ascend AI and Defend AI products with the introduction of capabilities for agentic security. The launch includes newly available “attack and defense agents” that can provide continuous security testing and automated enforcement, the company said. Through being trained on real-world exploits of AI agents, the new capabilities enable Straiker to offer the “first comprehensive solution for agentic AI threats,” the startup said.

WitnessAI

Founded: 2023

CEO: Rick Caccia

In an expansion to its Secure AI Enablement Platform, initially launched in October 2024, WitnessAI debuted two new products in August for bringing security to enterprise LLMs and AI-powered applications. The launch includes Witness Attack, a tool for automated red-teaming that can detect vulnerabilities in LLMs prior to deployment, the company said. The second new product, Witness Protect, provides “next-generation” AI firewall capabilities such as “behavioral runtime defense,” which can protect against AI model attacks including prompt injection and jailbreaking, according to WitnessAI.