---
title: "What 100 AI security vendors sell in 2026: 13 capabilities counted"
canonical: https://pharosone.ai/knowledge/ai-security-market-2026
publisher: PharosOne (Agentic Mind, Inc.)
updated: 2026-10-23
license: "Open for reuse: crawling, copying, quoting, summarising and model training permitted without asking; citation of the title and canonical URL requested (see https://pharosone.ai/terms#reuse)"
cite_as: "\"What 100 AI security vendors sell in 2026: 13 capabilities counted\". PharosOne, 2026-10-23. https://pharosone.ai/knowledge/ai-security-market-2026"
---

# What 100 AI security vendors sell in 2026: 13 capabilities counted

> A sample of 100 AI security vendors, 77 of them exhibitors at two European events: who sells guardrails, who tests agents via tools and MCP, who was acquired.

URL: https://pharosone.ai/knowledge/ai-security-market-2026 · Category: Reports · Published: 2026-10-22 · Updated: 2026-10-23 · Author: PharosOne Research · Reviewed by: PharosOne editorial

*PharosOne sells pentests, AI agent testing and a runtime guard, and wrote this article. PharosOne is not one of the 100 vendors counted.*

**In short.** We checked a sample of 100 vendors that sell security for AI systems and agents, 77 of them exhibitors at two European events, on 21–22 October 2026, against 13 capabilities. The sample is no market census, and the counts below describe it alone. Runtime guardrails are the most crowded category: 36 vendors make them a core capability and 28 more offer them in part. Testing agents through the tools and MCP servers they call is core for 11 vendors, and a human red team service for AI systems is core for 3. Nine vendors in the sample were acquired between July 2025 and September 2026, most of them by security or networking platforms.

## Key facts

- Updated: 23 October 2026.
- Sample: 100 vendors selling security for AI systems and agents. 77 come from the exhibitor lists of Web Summit Lisbon 2026 (67) and Barcelona Cybersecurity Congress 2026 (10); 23 were added by web research: 20 on 21 October as notable players in AI red teaming, agent and MCP security testing, guardrails and AI security posture management, chosen by evidence of market presence in the previous 12 months (funding, named customers, analyst mentions), and 3 on 22 October (Aikido Security, Intruder, Snyk) because they sell AI pentests with published prices. The exhibitor lists favour European and early-stage vendors. Protect AI and Palo Alto Networks are counted separately; Palo Alto acquired Protect AI and both now sell Prisma AIRS, so the 100 entries describe 99 distinct product lines. PharosOne is not in the sample.
- Method: AI research agents read each vendor's own product pages, documentation, pricing pages, changelogs, GitHub and press against 13 capabilities, and recorded each mark (core, partial, none or unknown) with its evidence, source URL and date seen in a fixed schema. "None" means the pages describe the vendor's products and the capability is absent from them; "unknown" means the public pages were too thin to tell, and the agents were told to mark unknown rather than guess. A separate fact-check agent re-checked 56 claims about acquisitions, funding rounds and events for 18 vendors: 49 confirmed, 7 corrected before publication, none unconfirmed. The capability marks rest on the research agents' reading of each source. Checked on 21–22 October 2026.
- Names: vendors are named for public facts with a source link: acquisitions, and the vendors marked core for testing agents through tools and MCP or for a human red team, each with the page the mark rests on.
- Data: all 100 vendors as CSV, [ai-security-vendors-2026-10.csv](/data/ai-security-vendors-2026-10.csv), with category, each capability mark and the page it rests on, and the date checked; CC BY 4.0. Method and limitations are in the [README](/data/README.md).
- Biggest gaps: testing agents through their tools (11 core) and human red teaming of AI systems (3 core).
- Next step: read [What an AI agent pentest costs in 2026](/knowledge/ai-agent-pentest-cost-2026) for published prices by vendor.

## What does the market sell?

| Capability | Core | Partial | None | Unknown |
|---|---:|---:|---:|---:|
| Runtime guardrails | 36 | 28 | 31 | 5 |
| Compliance evidence | 29 | 47 | 4 | 20 |
| AI inventory and shadow AI discovery | 24 | 21 | 26 | 29 |
| Automated LLM red teaming | 21 | 9 | 68 | 2 |
| Prompt injection testing | 21 | 9 | 62 | 8 |
| Web application pentest | 19 | 4 | 70 | 7 |
| Agent identity and authorisation | 18 | 26 | 31 | 25 |
| DLP for LLM traffic | 17 | 22 | 33 | 28 |
| Monitoring and observability | 17 | 50 | 15 | 18 |
| Testing agents through tools and MCP | 11 | 17 | 64 | 8 |
| AI security posture management | 9 | 17 | 33 | 41 |
| Model supply-chain scanning | 7 | 13 | 60 | 20 |
| Human red team service for AI | 3 | 16 | 65 | 16 |

"Core" means the capability is a named product or a central claim on the vendor's own pages; "partial" means it appears as a feature of something else or in limited form; "unknown" means the public pages did not say. By category, AI security platforms are the largest group (24 vendors), followed by AI compliance tools (17), other vendors (17), general application security vendors adding AI features (15), LLM guardrails (12), AI pentest vendors (8) and AI red team vendors (7).

## Where is the market crowded?

Runtime guardrails and compliance tooling. 64 vendors offer guardrails in core or partial form. 76 offer some form of compliance evidence, and 4 of the 100 mention AIUC-1, a certification standard written for AI agents, in the evidence we found.

## Where are the gaps?

**Testing agents through what they touch.** An agent's risk sits in its tools, its MCP servers, its memory and the documents it reads; the OWASP Agentic Top 10 lists tool misuse and memory poisoning among the main risks ([OWASP](https://genai.owasp.org/resource/owasp-top-10-for-agentic-applications-for-2026/)). 11 vendors make testing through tools and MCP a core capability and 17 offer it in part.

**People.** Human red teaming of AI systems is core for 3 vendors and partial for 16. 22 vendors make offensive testing of AI systems a core capability: the 21 marked core for automated LLM red teaming (all 11 marked core for testing agents through tools and MCP are among them) plus 1 marked core for a human red team service with prompt injection testing. 10 of those 22 also offer a human red team in core or partial form.

**The full scope in one place.** 2 of the 100 offer offensive testing of AI systems, a human red team, a classic web pentest and compliance evidence, each at least in partial form; none offers all four as core. This combination mirrors what PharosOne sells, so read it with the disclosure at the top. A buyer who needs a pentest report for SOC 2 or DORA and evidence about the agents on the same stack may end up with two or three vendors.

## Which vendors test agents through their tools and MCP servers?

11 vendors in the sample make it a core capability, as their own pages describe it. The list is ordered by name, with the page each mark rests on:

| Vendor | Page the mark rests on |
|---|---|
| Cisco AI Defense | [Cisco blog: AI Defense Explorer](https://blogs.cisco.com/ai/introducing-cisco-ai-defense-explorer) |
| Enkrypt AI (acquired by Anaconda) | [Agent red teaming](https://enkryptai.com/product/agent-red-teaming) |
| ForteShield AI | [Releases](https://forteshield.ai/releases) |
| Lasso Security | [AI red teaming](https://www.lasso.security/platform/ai-red-teaming) |
| Mindgard | [Automated AI red teaming](https://mindgard.ai/automated-ai-red-teaming) |
| Pillar Security | [Red teaming and risk detection](https://www.pillar.security/platform/red-teaming-risk-detection) |
| Promptfoo (joining OpenAI) | [MCP security testing](https://www.promptfoo.dev/docs/red-team/mcp-security-testing/) |
| Repello AI | [Blog: ARTEMIS and MCP](https://repello.ai/blog/claude-code-ai-red-teaming) |
| Snyk | [Evo launch](https://snyk.io/news/snyk-launches-evo-continuous-offensive-security) |
| SPLX (acquired by Zscaler) | [Zscaler continuous automated red teaming](https://www.zscaler.com/products-and-solutions/continuous-automated-red-teaming) |
| Straiker | [Ascend AI](https://www.straiker.ai/products/ascend-ai) |

A human red team service for AI systems is core for 3 vendors: [Gray Swan AI](https://www.grayswan.ai/solutions/ai-red-teaming), [Kiggu](https://kiggu.io/solutions) and [Pentestiverse](https://pentestiverse.com/services/llm-red-teaming.html). Pages were read on 21–22 October 2026; a vendor whose page has changed since may be marked differently today.

## Is the market consolidating?

Yes. 9 of the 100 vendors were acquired by larger security, networking, platform or AI companies between July 2025 and September 2026, among them Protect AI, which Palo Alto Networks now sells as part of Prisma AIRS. The acquired products are now sold as modules of the buyer's platform.

| Vendor | Acquired by | Announced or closed | Source |
|---|---|---|---|
| Protect AI | Palo Alto Networks | completed 22 July 2025 | [Palo Alto Networks](https://investors.paloaltonetworks.com/news-releases/news-release-details/palo-alto-networks-completes-acquisition-protect-ai/) |
| Aim Security | Cato Networks | announced 3 September 2025 | [CyberScoop](https://cyberscoop.com/cato-networks-acquires-ai-security-startup-aim-security/) |
| Prompt Security | SentinelOne | September 2025 | [SentinelOne 10-Q](https://www.sec.gov/Archives/edgar/data/1583708/000158370825000159/s-20251031.htm) |
| Lakera | Check Point | announced 16 September 2025 | [Check Point](https://www.checkpoint.com/press-releases/check-point-acquires-lakera-to-deliver-end-to-end-ai-security-for-enterprises/) |
| CalypsoAI | F5 | September 2025 | [F5 10-K](https://www.sec.gov/Archives/edgar/data/1048695/000104869525000157/ffiv-20250930.htm) |
| SPLX | Zscaler | announced 3 November 2025 | [Zscaler](https://www.zscaler.com/press/zscaler-secures-enterprise-ai-lifecycle-acquisition-innovative-ai-security-pioneer-splx) |
| Promptfoo | OpenAI | agreement announced 9 March 2026 | [Promptfoo](https://www.promptfoo.dev/blog/promptfoo-joining-openai/) |
| Enkrypt AI | Anaconda | 4 August 2026 | [Anaconda](https://www.anaconda.com/press/anaconda-acquires-enkrypt-ai) |
| Haize Labs | Beacon Software | announced 17 September 2026 | [Business Wire](https://www.morningstar.com/news/business-wire/20260917239094/beacon-acquires-haize-labs-to-power-ai-for-the-real-economy) |

For a buyer this raises one question: is the tester independent of the platform it tests?

## How are prices shown?

30 of the 100 vendors show at least one price on their own pages or a review site; the other 70 publish no price and sell through a demo, a quote or a sales call. 33 offer something free: a free tier, a free trial or a free open-source tool. Published prices by vendor and segment are in [What an AI agent pentest costs in 2026](/knowledge/ai-agent-pentest-cost-2026).

## Where does PharosOne sit?

PharosOne is not in the sample. It sells a [human pentest](/knowledge/human-pentest), an [AI-run assessment](/knowledge/ai-agent-security-assessment) and the runtime guard [Pharos Radar](/knowledge/pharos-radar); see [which product fits which situation](/knowledge/which-product-do-i-need).

## FAQ

### How many vendors sell security for AI agents in 2026?

This review covers 100 vendors selling security for AI systems and agents, checked on 21–22 October 2026. 22 of them sell offensive testing of AI systems as a core capability. The sample is not the whole market: 77 vendors come from two 2026 conference exhibitor lists and 23 were added by web research.

### What is the most common AI security capability?

Runtime guardrails. 36 of the 100 vendors make them a core capability and 28 more offer them in part.

### How many vendors combine automated testing with a human red team?

10 of the 22 vendors that sell offensive testing of AI systems also offer a human red team service, in core or partial form. Human red teaming is a core capability for 3 of the 100.

### Is the market consolidating?

Nine of the 100 vendors were acquired between July 2025 and September 2026, among them Lakera by Check Point, Protect AI by Palo Alto Networks, Prompt Security by SentinelOne, CalypsoAI by F5 and Promptfoo by OpenAI.

### Is PharosOne in the sample?

No. PharosOne compiled the review and is not one of the 100 vendors counted.

## Sources

1. [PharosOne AI security vendor map, 100 vendors, 13 capabilities (compiled 21–22 October 2026), CSV, CC BY 4.0](https://pharosone.ai/data/ai-security-vendors-2026-10.csv). PharosOne Research, 2026-10-23.
2. [Palo Alto Networks completes acquisition of Protect AI](https://investors.paloaltonetworks.com/news-releases/news-release-details/palo-alto-networks-completes-acquisition-protect-ai/). Palo Alto Networks, 2025-07-22.
3. [Cato Networks acquires AI security startup Aim Security](https://cyberscoop.com/cato-networks-acquires-ai-security-startup-aim-security/). CyberScoop, 2025-09-03.
4. [SentinelOne quarterly report (10-Q): acquisition of Prompt Security](https://www.sec.gov/Archives/edgar/data/1583708/000158370825000159/s-20251031.htm). SentinelOne, U.S. SEC, 2025-10-31.
5. [Check Point acquires Lakera to deliver end-to-end AI security for enterprises](https://www.checkpoint.com/press-releases/check-point-acquires-lakera-to-deliver-end-to-end-ai-security-for-enterprises/). Check Point Software Technologies, 2025-09-16.
6. [F5 annual report (10-K): acquisition of CalypsoAI](https://www.sec.gov/Archives/edgar/data/1048695/000104869525000157/ffiv-20250930.htm). F5, Inc., U.S. SEC, 2025-09-30.
7. [Zscaler secures the enterprise AI lifecycle with acquisition of SPLX](https://www.zscaler.com/press/zscaler-secures-enterprise-ai-lifecycle-acquisition-innovative-ai-security-pioneer-splx). Zscaler, 2025-11-03.
8. [Promptfoo is joining OpenAI](https://www.promptfoo.dev/blog/promptfoo-joining-openai/). Promptfoo, 2026-03-09.
9. [Anaconda acquires Enkrypt AI](https://www.anaconda.com/press/anaconda-acquires-enkrypt-ai). Anaconda, Inc., 2026-08-04.
10. [Beacon acquires Haize Labs to power AI for the real economy](https://www.morningstar.com/news/business-wire/20260917239094/beacon-acquires-haize-labs-to-power-ai-for-the-real-economy). Beacon Software, Business Wire, 2026-09-17.
11. [OWASP Top 10 for Agentic Applications 2026](https://genai.owasp.org/resource/owasp-top-10-for-agentic-applications-for-2026/). OWASP GenAI Security Project, 2025-12-09.
12. [What an AI agent pentest costs in 2026](https://pharosone.ai/knowledge/ai-agent-pentest-cost-2026). PharosOne Research, 2026-10-22.

Source: PharosOne, "What 100 AI security vendors sell in 2026: 13 capabilities counted", https://pharosone.ai/knowledge/ai-security-market-2026, updated 2026-10-23.

---

PharosOne is a product of Agentic Mind, Inc., an independent AI agent security company. Site map for agents: https://pharosone.ai/llms.txt