AI penetration testing companies: who does what
AI penetration testing means two things: testing done with AI agents, and testing of AI systems such as LLM apps. This list covers the first meaning: providers whose delivery is automated, or whose type is autonomous or AI-assisted. Providers that test AI systems are on a separate list. The table is A to Z, not ranked, and each cell carries the date we checked it.
AI penetration testing is a mixed term, and this page keeps the two meanings apart. This list covers companies that test with AI: autonomous and AI-assisted platforms and services. It does not cover providers that test AI systems, such as LLM apps and agents. For those, open the AI red teaming companies list. The table is A to Z, not ranked, and no provider pays for position.
21 of 21 providers shown (67 in the full table). Sorted A to Z, not ranked. Checked up to 9 November 2026.
| Provider | Delivery | Web apps | APIs | Cloud | Network | Mobile | AI apps | Published price | Turnaround | Retest | Sample report |
|---|---|---|---|---|---|---|---|---|---|---|---|
| AikidoApplication security platform, Multiple (United States, Belgium, United Kingdom) | automated | Yes | Yes | ? | Yes | Yes | Yes | $4,000 per assessment +1 more | TYPICAL Pentest: same-day results. The AI pentest page also says a… | Free retesting for up to 6 months, including checks for bypasses of… | Public |
| DetectifyApplication security platform | automated | Yes | Yes | ? | ? | ? | ? | from EUR 0 per year (annual platform fee) +4 more | Not published | Not stated on the pricing page. | not published |
| EquixlyAutonomous testing platform | automated | Yes | Yes | ? | ? | ? | Yes | Custom quote | Not published | Revalidates whenever applications and AI systems change; no separate… | not published |
| EscapeApplication security platform, France | automated | Yes | Yes | ? | Yes | ? | Yes | Custom quote | Continuous AI pentesting: results in minutes on merge, deploy or… | Open issues are re-checked; external network pentesting is followed… | not published |
| Excession LabsPublisherSecurity lab (AI-assisted testing) | hybrid | Yes | Yes | ? | ? | ? | Yes | Custom quote | Not published | Retest of fixed findings is offered; terms are set in each quote. | Public |
| GiskardAI-assisted testing service, France | hybrid | ? | Yes | ? | ? | ? | Yes | $0 no billing unit given | Not published | Giskard says it re-runs every test until the fix is confirmed. | not published |
| HadrianAutonomous testing platform, Netherlands | hybrid | Yes | ? | ? | ? | ? | ? | From EUR 1,250 per test, excluding VAT | Most tests finish within 24 to 48 hours; same-day start possible. | Not published | not published |
| Horizon3.ai (NodeZero)Autonomous testing platform, United States | automated | Yes | Yes | Yes | Yes | ? | ? | $15 per asset, per test (12-month term) +4 more | Not published as a time. WebApp page says tests can be set up… | Findings include fix actions with one-click verify. On Core, Pro and… | not published |
| IntruderApplication security platform, United Kingdom | automated | Yes | ? | ? | ? | ? | ? | $3,500 per test +2 more | Setup in minutes; results within hours; the pricing page says… | Unlimited retesting on the single test and the 4-pack. Packs expire… | not published |
| NoveeAutonomous testing platform | automated | Yes | Yes | ? | ? | Yes | Yes | Custom quote | Not published. Testing is described as continuous. | Retests run automatically after fixes, and on new deployments, code… | not published |
| ParameterAI-assisted testing service | automated | Yes | Yes | ? | ? | ? | ? | Custom quote | First findings in 24 hours; most runs finish the same day they start. | Free re-testing is listed for Single App and inherited by higher… | Public |
| PenligentAI-assisted testing service | automated | Yes | ? | ? | Yes | ? | ? | $0 per month +2 more | What takes humans a week, Penligent takes an hour (homepage tagline). | Pro includes task re-runs, scheduled recurring runs and risk… | Public |
| PenteraAutonomous testing platform | hybrid | Yes | Yes | Yes | Yes | ? | ? | Custom quote | Not published | Remediation is followed by re-testing; the FAQ says it revalidates… | not published |
| PentiAI-assisted testing service | hybrid | Yes | Yes | Yes | Yes | Yes | Yes | $20 per month ($216 per year billed annually) +5 more | Pentests done in hours, not months (tagline). On the free first… | Every plan includes free unlimited retests. | not published |
| RunSybilAutonomous testing platform | automated | Yes | Yes | Yes | ? | ? | ? | Custom quote | Not published | No retest policy; the platform re-evaluates on every deployment and… | on request |
| SPLXApplication security platform | automated | ? | Yes | ? | ? | ? | Yes | Custom quote | Not published | No retest policy; testing is described as continuous. | not published |
| StrixAutonomous testing platform | automated | Yes | Yes | Yes | ? | ? | ? | $29 per seat per month | Not published | The platform retests its own generated fixes; no retest policy or… | not published |
| Terra SecurityAI-assisted testing service | hybrid | Yes | ? | ? | Yes | ? | Yes | Custom quote | Vendor says tests drop from 4-6 weeks to 2-4 hours versus… | Not published | not published |
| TurboPentestAI-assisted testing service | automated | Yes | Yes | ? | Yes | ? | Yes | $99 per credit (one pentest) +4 more | All tiers typically complete in a few hours. | A retest is a full new pentest at the same per-target price; no… | Public |
| XBOWAutonomous testing platform | automated | Yes | Yes | ? | ? | ? | ? | Custom quote | First validated findings arrive in hours; testing is continuous or… | Retests after a fix are described as free and unlimited. | not published |
| ZeroThreatAutonomous testing platform | automated | Yes | Yes | ? | ? | ? | ? | $100 per target per month +2 more | Not published | Professional: unlimited retesting. Pay Per Scan: unlimited retesting… | not published |
No provider matches every filter. Remove a filter to see more.
"?" means the provider does not say on its public pages. It does not mean "no". Prices use each provider's own unit: a test, a year, a month or credits. Do not compare them as equal units. How we collect and check this data.
Each cell comes from the provider's own pages, with the date we checked. "Not published" means we found no statement.
What does this list include?#
A provider qualifies when its dataset record has one of three markers: delivery is "automated", or its type is "autonomous" or "AI-assisted". The rule is mechanical. We do not judge whether a vendor's AI claim is true.
The Delivery column shows the split, using our methodology definitions:
| Delivery value | What it means in our dataset |
|---|---|
| automated | Software or AI agents run the tests, and any human step is limited or not described |
| hybrid | People and AI agents or a crowd work together, or the provider sells both |
| human-led | People run the tests. A provider appears on this list only through its type |
Some vendors sell both an AI product and a human-led service. The row shows the main product. Open the profile for the rest.
Who does it help?#
Teams that release often and want repeated tests may look at automated and hybrid providers. Teams that need a report for an auditor or a customer, or that have complex business logic, should read the sample report before they choose. The guide on AI penetration testing services compares prices, limits and retest terms in prose. Use this page to scan the data and that guide to read the detail.
If your question is whether AI testing replaces a person, read AI vs human pentesting.
What to ask providers#
Send the same questions to every provider and keep the written replies.
- Who checks a finding before it reaches me? Name the role.
- What does the agent do on its own, and what needs approval?
- Can the agent change data? Under which limits?
- Which scope does a test cover: web app, API, authenticated areas, network?
- How does the report prove a finding: request and response, or a score?
- Is a retest included, for how long and on which findings?
- Where is our data stored, and does a model vendor receive it?
- Can we see a sample report? The table shows which providers publish one.
Price drivers#
Prices on this list use different units, so a ranking by price would mislead. Published units include per test, per asset, per credit and per month, and many providers publish nothing. What changes a price:
- App size and number of user roles
- Number of assets or targets
- White-box access, such as a code repository link
- Whether a human review or a retest is included
- Subscription term versus a single test
For a wider view of units and amounts, read penetration testing cost.
How to read the table#
Start with Delivery and Type, then read the scope columns: Web, API, Cloud and AI apps. "Unknown" means we found no statement. The AI apps column shows whether a provider also says it tests AI systems, so a row can sit on both lists.
Published price shows the first listed price and its unit, or the pricing model when no price is listed. Turnaround repeats the provider's wording. Providers measure different events, for example time to start, test duration or time to first findings, so compare the meaning, not the number. Retest shows the provider's retest terms. Sample report shows whether a report is public, on request or not published.
Limits of the list#
- The list shows stated products, not tested performance. We do not run the tools.
- Vendor claims about AI are the vendor's own. We record them with a source.
- Marketing pages change. The date on each profile is the date we last checked.
- A provider with no row may still fit your need. Missing providers can be added through List your company.
Where to go next#
Compare providers across lists in the provider directory or the lists index. For test depth and limits, read automated pentesting tools.
Questions
What is the difference between AI-driven pentesting and AI red teaming?
AI-driven pentesting uses AI agents to test normal systems such as web apps and networks. AI red teaming tests AI systems, such as an LLM app, for prompt injection and data leaks. A provider can do one, both or neither. Check which meaning each vendor page uses.
Does an AI pentest include a human reviewer?
It depends on the provider. Some say a human reviews findings, some sell the test as fully automated, and some offer a human-led option. The Delivery column shows what each provider states. Ask who checks a finding before you see it.
Can an AI pentest replace a human-led pentest?
The dataset cannot say. Providers differ in scope, depth and what the report proves. If an auditor or customer asks for a human-led test, ask them first. Compare sample reports and retest terms before you decide.
Why are some prices missing in the table?
Not every provider publishes a price. Where one does, the table shows the first listed price with its unit, and units differ: per test, per asset, per credit or per month. Open the provider profile for the full price list and what each price covers.
Sources
- PentestCompare provider dataset (list: ai-penetration-testing-companies) (checked 9 November 2026)
- PentestCompare methodology (checked 9 November 2026)
Compare like with like
An automated scan, a manual test and annual credits are different units. Check scope, delivery and retest terms before price.