Verified
Reviewed byOleh KemOleh Kem
Plans checked19 / 5 vendors
“Contact sales”5/19
Median entry$36/mo
Entry range$12-60
Publish a price2/5
Free tier1/5
Leader87 Microsoft Purview

AI security software compared: pricing, runtime guard and governance

Two markets hide under one label: software that guards a live AI app, and software that governs the model behind it, billed in units that never line up. Only two of the five vendors name a price you can use, so for the rest a marketplace estimate, where one even exists, is as close to a number as you get.

How much does AI security software cost?

The middle of this small market is about $36 a month, though that median stands on only two public prices, $12 and $60, checked July 29, 2026. The rest sell by quote. What you protect, the running app or the model behind it, decides which budget you are in.

  • Only two of the five vendors put a number in public, checked July 29, 2026. The rest name none until you reach a sales team.
  • The public monthly prices are $12 and $60. With only two of them, the $36 median is a midpoint, not a going market rate.
  • The two halves of the market meter differently. Guard tools charge for traffic through a gateway; governance tools charge per AI use case or asset. The units do not convert.
  • Five of the 19 plans on the record carry no published price. Most of the priced ones belong to a single usage-metered vendor, which makes the market look more open than it is.
  • Microsoft Purview tops the disclosure ranking at 87 of 100; Arthur AI follows at 70.

Ranked by a transparency score: pricing transparency 60%, user satisfaction 40%. Capability is not scored. It is the condition grid below. Prices are read from vendor pricing pages and re-checked per product on the dates shown. ComparEdge sells no AI security software and takes no payment for placement. How the score is built.

01 / 06

AI security software ranked: runtime guard, shadow AI and PII redaction

The order here is a disclosure order, not a price order: whoever reveals the most about what they charge ranks first. The capability columns are a runtime guard, shadow-AI discovery, PII redaction, AI governance, model monitoring, and a SIEM feed, and they fall into two groups. The first set protects a live app; the rest watch the model and document it. An empty cell records only that the vendor left that feature off its pages when we last looked.

Sorted by transparency scorePriced tiers 14 / 19Full disclosure 1 / 5
How to read this table
01Microsoft PurviewNo free tierAll tiers pricedRuntime guard not on the recordShadow AI listedPII redaction listedAI governance listedModel monitoring not on the recordSIEM feed listed$12/u87Alternatives to Microsoft Purview
02Arthur AIFree tier1 of 3 tiers unpricedRuntime guard not on the recordShadow AI not on the recordPII redaction not on the recordAI governance listedModel monitoring listedSIEM feed not on the record$60flat70Alternatives to Arthur AI
03Credo AINo published priceNo free tier0 of 1 tiers pricedRuntime guard not on the recordShadow AI not on the recordPII redaction not on the recordAI governance listedModel monitoring listedSIEM feed not on the recordSales only38Alternatives to Credo AI
04Lasso SecurityMarketplace estimatefrom $20,000 / yrNo free tier0 of 2 tiers pricedRuntime guard listedShadow AI listedPII redaction listedAI governance not on the recordModel monitoring not on the recordSIEM feed listedSales only0Alternatives to Lasso Security
05HiddenLayerNo published priceNo free tier0 of 1 tiers pricedRuntime guard listedShadow AI not on the recordPII redaction not on the recordAI governance not on the recordModel monitoring listedSIEM feed listedSales only0Alternatives to HiddenLayer
GrantedSome tiers sealedNot on the recordCheapest paid seatFlat account fee, not per seat* score is user satisfaction alone: the vendor publishes no pricesScore ranks pricing transparency and user ratings, not capability. Capability is the grid.
5 vendorsMedian entry $36Coverage span 4 → 2 of sixSealed tiers 5 / 19

Scroll the console sideways to reach the remaining conditions.

02 / 06

Every vendor's plans, verdicts and the date we checked each price

Sequence follows disclosure, highest first, not headline price and not popularity. Each row opens to the vendor's full plan record, an expert read, and links through to our pricing and alternatives pages. The two that price in the open hold the leading spots.

Transparency scorePricing transparency 60%User satisfaction 40%

01

Microsoft Purview, the data governance and compliance suite built into Microsoft 365 and Azure
Microsoft Purview

87Disclosure$12Seat / mo, billed yearly

Enterprises whose stack is already Microsoft 365 and Azure, and who want to govern the data their AI touches inside tools they run every day. Purview handles shadow-AI discovery, PII redaction and audit through licences many of them already hold. Its add-ons meter separately, and the bill gets complicated fast.

Critical gapThe platform requires significant administrative configuration overhead for initial policy deployment and schema mapping.

Plan table and expert take

Microsoft Purview: expert take

Purview rides Microsoft 365 and Azure licensing many buyers already own, so its entry price misleads. The real bill lands on eleven separate meters, from per-asset-day to per-ten-thousand-requests. It reacts after a breach rather than blocking a live prompt, closer to data governance than a runtime guard.

Where Microsoft Purview holds up

  • Natively protects data used by Microsoft 365 Copilot and other GenAI.
  • Deepest integration with M365, Teams, SharePoint, and Azure services.
  • Unified data classification, labeling, and protection across the ecosystem.
  • Strong Insider Risk Management and Communication Compliance features.
  • Advanced eDiscovery (Premium) streamlines legal holds and review.

4.5CE scoreG2 4.4 · 32 reviewsCapterra 4.3Founded 2021Verified July 17, 2026

14 plans, as published
PlanMonthlyAnnual
Microsoft 365 E5Not published$60
Microsoft 365 E5 (no Teams)Not published$51.45
Microsoft Purview SuiteNot published$12
At Rest Protection$0.02 / asset/dayNot published
In Transit Protection$0.50 / 10K requestsNot published
Insider Risk Management$25 / DSPUNot published
Data Security Investigations$5 / compute unit/hrNot published
On-Demand Classification$20 / 10K assetsNot published
Audit$0.01 / 1K recordsNot published
Data Lifecycle Management$0.0002 / 1K messages/dayNot published
eDiscovery$0.67 / GB/dayNot published
Communication Compliance$0.30 / 1K text recordsNot published
Unified Catalog$0.02 / asset/dayNot published
Data Governance Processing Unit$15 / DGPUNot published
See pricingRead the reviewAlternatives
Vendor pricing page
02

Arthur AI, the model monitoring platform that tracks bias, drift, and LLM performance in production
Arthur AI

70Disclosure$60Flat / mo

Data and ML groups that watch models in production for bias, drift and LLM quality. Arthur is the single vendor you can buy straight from its site, which makes it the easiest place to start. It watches models rather than blocking them, closer to observability than a firewall.

Critical gapThe system lacks offline deployment capabilities for air-gapped environments.

Plan table and expert take

Arthur AI: expert take

Arthur is the one vendor you can buy without a sales call: a free tier watches four use cases, the paid step lifts that to a hundred. It monitors models for bias and drift instead of blocking attacks, so treat it as observability, not a firewall.

Where Arthur AI holds up

  • Best-in-class bias and fairness detection
  • Arthur Bench provides rigorous LLM evaluation
  • Strong in regulated industries (finance, healthcare)
  • Explainability features for compliance documentation

4.6CE scoreG2 4.5 · 10 reviewsFounded 2018Verified July 8, 2026

3 plans, as published
PlanMonthlyAnnual
FreeFreeFree
Premium$60Not published
EnterpriseContact sales
See pricingRead the reviewAlternatives
Vendor pricing page
Below this line, nobody posts a figure you can plan against. The number arrives from a sales team, after the demo and the security review.
03

Credo AI, the AI governance platform that maps model risk to NIST and the EU AI Act
Credo AI

38DisclosureSales onlyNo price

Organizations standing up a formal AI governance program mapped to the EU AI Act, NIST and internal policy. Credo turns model and risk artifacts into audit-ready reports and prices by the count of governed use cases. It does nothing at runtime, so it sits beside a firewall, not in place of it.

Plan table and expert take

Credo AI: expert take

Credo prices by the count of AI use cases you put under governance, plus an overage for each one past the plan. It maps models to the EU AI Act and NIST and generates audit files. At runtime it does nothing, which is why governance buyers run it alongside a guard, never instead of one.

Where Credo AI holds up

  • Deep alignment with EU AI Act, NIST RMF, ISO 42001
  • Workflow-driven AI policy management
  • Generates audit-ready compliance documentation
  • Vendor AI risk assessment capabilities

5.0CE scoreG2 4.9 · 6 reviewsFounded 2020Verified July 8, 2026

1 plans, as published
PlanMonthlyAnnual
EnterpriseContact sales
See pricingRead the reviewAlternatives
Vendor pricing page
04

Lasso Security, the runtime firewall that inspects LLM prompts and responses for injection and data leakage
Lasso Security

0DisclosureSales onlyNo price

Anyone putting a customer-facing LLM app into production and needing a guard in front of it. Lasso reads prompts and responses in real time to catch injection and data leakage, and works with any model API. It sells enterprise-only, so a small team finds no self-serve door.

Critical gapSystem performance degrades significantly during complex, high-volume data processing tasks.

Plan table and expert take

Lasso Security: expert take

Lasso sells two priced things under one name: a staff-chatbot product and an LLM gateway, quoted separately on its marketplace listing, with a per-unit charge stacked on top of the annual commit. Neither has a self-serve tier, so it is an enterprise buy.

Where Lasso Security holds up

  • Real-time LLM firewall with low latency
  • Strong focus on prompt injection and jailbreak detection
  • Founded by experienced cybersecurity professionals
  • Model-agnostic: works with any LLM API

Founded 2023Verified July 8, 2026

2 plans, as published
PlanMonthlyAnnual
Lasso for EmployeesContact sales
Secured Gateway for LLMsContact sales
See pricingRead the reviewAlternatives
Vendor pricing page
05

HiddenLayer, the model-layer security platform that guards machine learning weights and pipelines
HiddenLayer

0DisclosureSales onlyNo price

Security teams protecting the model itself, its weights and training pipeline, from theft and adversarial attacks. HiddenLayer works at the model layer rather than the prompt layer, which is the gap most vendors here leave open. It fits organizations guarding AI as intellectual property more than teams filtering everyday chat.

Critical gapThe platform requires active integration into existing model pipelines to function as a runtime monitor.

Plan table and expert take

HiddenLayer: expert take

HiddenLayer guards the model itself, its weights and training pipeline, the layer most tools here ignore. It publishes no price at all. The large figure on its marketplace listing is a ceiling that routes buyers to a private offer, not a rate anyone actually pays.

Where HiddenLayer holds up

  • Unique focus on ML model-level security (weights, pipelines)
  • Detects adversarial attacks at inference time
  • Supply chain security for ML models
  • Fills a gap most AI security tools ignore

Founded 2022Verified July 8, 2026

1 plans, as published
PlanMonthlyAnnual
AISec PlatformContact sales
See pricingRead the reviewAlternatives
Vendor pricing page
03 / 06

Compare any two AI security vendors: plans, limits and our score

vs

What the records say

For 6 seats, Arthur AI bills $60 / mo and Microsoft Purview bills $72 / mo, $12 / mo between them.

Arthur AI charges a flat account fee, so it undercuts Microsoft Purview from 6 seats up.

Microsoft Purview carries 4 of the 6 capability columns on the record; Arthur AI shows 2.

Users side with Arthur AI: 4.5 on G2 against 4.4 for Microsoft Purview.

Arthur AI runs a free tier to start on; Microsoft Purview does not.

Microsoft Purview prices everything it sells; Arthur AI leaves part of its lineup unpriced.

Pick Microsoft Purview for: Enterprises whose stack is already Microsoft 365 and Azure, and who want to govern the data their AI touches inside tools they run every day.

Pick Arthur AI for: Data and ML groups that watch models in production for bias, drift and LLM quality.

№ 01

Microsoft Purview

CE 87 · G2 4.4
Published plans, US$/mo
Microsoft 365 E5Not published
Data Security Investigations$5 / compute unit/hr
Data Governance Processing Unit$15 / DGPU
Team of 6$72 / mo

Verified July 17, 2026

№ 02

Arthur AI

CE 70 · G2 4.5
Published plans, US$/mo
FreeFree
Premium$60
EnterpriseContact sales
Team of 6$60 / mo

Verified July 8, 2026

Both price lists on the category axis

Microsoft Purview
Arthur AI

Where they differ

Only Microsoft Purview has on the record

  • Every tier priced
  • Shadow AI
  • PII redaction
  • SIEM feed

Only Arthur AI has on the record

  • $0 tier
  • Model monitoring
Full comparison: Microsoft Purview alternativesMicrosoft Purview pricingArthur AI pricing
04 / 06

AI security questions: what it costs, why prices are hidden, governance

How much does AI security software cost?

There is no single number, because the name covers two different products. The two vendors with a monthly price sit at $12 and $60. Everything else is annual and quoted. A runtime gateway or an enterprise governance program can reach into the hundreds of thousands a year once the traffic or the use-case count grows.

Why don't AI security vendors publish prices?

Most sell into security and compliance teams at large companies, a slow sale with scoping and a security review. A public price list does not fit that motion. It shows in the record: of the five here, only two name a figure you can act on. The rest treat price as the last thing you learn, after the demo.

What is the difference between AI security and AI governance?

AI security guards the running system. It watches prompts and responses for injection, data leakage or abuse, and acts in real time. AI governance sits above the model, tracking which use cases are approved and whether they meet the EU AI Act or NIST, and produces the paperwork to prove it. Same category page, two jobs and two buyers.

How is AI security software priced?

Two ways, and they do not compare. The guard tools meter traffic: the requests or units flowing through a gateway, sometimes on top of an annual commit. The governance tools meter use cases or assets under management, with an overage for each one past your plan. So a quote for one side tells you nothing about the cost of the other.

Can you buy AI security software without talking to sales?

Barely. Only one of the five, Arthur AI, lets you sign up and check out directly, with a free tier that covers four use cases and a paid step that lifts the ceiling to a hundred. Microsoft Purview you buy through Azure or a Microsoft 365 licence. Everyone else routes you to a sales team first.

Is Microsoft Purview enough to secure AI?

It depends what you mean by secure. Purview governs and classifies the data your AI touches, which is real coverage if you live inside Microsoft 365. It does not inspect a live prompt or block an injection as it happens, so a customer-facing LLM app still wants a runtime guard in front of it. Purview is the governance half, not the firewall.

What does an LLM security audit cost?

None of the five platforms here sell a one-off audit; that is a separate service market. Independent red-teaming or a security assessment of an LLM is a one-time project, not a subscription, and the third-party firms that sell it publicly price such work from a few thousand into the tens of thousands. What this page covers is ongoing protection, priced by the year or by the unit of traffic.

Do you need both a firewall and a governance tool?

Often, yes, because they do different jobs. A runtime firewall like Lasso or HiddenLayer inspects what goes into and out of a model as it runs. A governance platform like Credo or Arthur documents risk and maps it to regulation after the fact. One catches an attack in the moment; the other proves you were watching. Large deployments end up buying both.
Field note 01

The two meters under an AI security bill

The two halves of this market do not just protect different things. They meter different things. A guard like Lasso bills for traffic: the requests or units running through a gateway, often on top of a yearly commit. A governance platform like Credo bills for use cases or assets under management, with an overage for each one past the plan.

That is why a single AI security budget number means little until you name the half. Priced per request, the cost tracks how busy your app is. Priced per use case, it tracks how many AI projects you are willing to govern. The two curves cross at different points, and no quote from one side predicts the other.

Field note 02

Reading an AI security price when the vendor won't print one

Most of the field publishes nothing, so buyers fall back on marketplace listings. Treat those with care. A round, very large number on an AWS Marketplace page is usually a ceiling, a placeholder that sends every real deal into a negotiated private offer. It is not the price you would pay.

The prices that are real here are the ones a vendor prints on its own pages, and there are only two of them. For everyone else, the useful question is not the headline but the unit. Ask what one seat, one use case, or ten thousand requests costs, and whether the annual commit carries an overage on top. That is the point where the real number takes shape.

The verdict on AI security softwareSigned review · Updated
Oleh KemOleh KemFounder & Lead AnalystComparEdge Editorial
ProfileORCIDMethod

Decide which problem you are buying for: guarding a deployed app is one product, governing models and data is another, and vendors rarely do both well. Public prices are the exception here; almost everything is a scoped conversation.

Treat capability grids as claims to verify in a proof of concept, not facts to buy on. Vendors who put a number on the page sit above vendors who put a form there.

MethodEvery price on this page is read from the vendor's own pricing page: 19 plans across five vendors, last verified .
DisclosureCollection is tool-assisted; every verdict is written and signed by a human analyst.
05 / 06
Not sure which? Answer one, take a shortlist.Pick the line that sounds like your team. Each one opens the vendor built for it.
Enterprises whose stack is already Microsoft 365 and Azure, and who want to govern the data their AI touches inside tools they run every dayMicrosoft PurviewData and ML groups that watch models in production for bias, drift and LLM qualityArthur AIOrganizations standing up a formal AI governance program mapped to the EU AI Act, NIST and internal policyCredo AIAnyone putting a customer-facing LLM app into production and needing a guard in front of itLasso SecuritySecurity teams protecting the model itself, its weights and training pipeline, from theft and adversarial attacksHiddenLayer

Still weighing it? Answer 5 questions, take a shortlist.

06 / 06

Read next: cost guides for AI security software, plus related categories

Related categoriesData ObservabilityCompliance AutomationEndpoint Security (XDR/EDR)Customer SupportLarge Language ModelsHR & People OpsCompare two vendorsMicrosoft Purview vs Arthur AIArthur AI vs Credo AICredo AI vs Lasso SecurityLasso Security vs HiddenLayerStill unsureAnswer 5 questions → a shortlistAll categories

How this review is made. Prices are read from vendor pricing pages and re-checked on the dates shown against each product. Condition columns reflect the feature set recorded on the vendor’s own pages on that date. ComparEdge sells no AI security software and takes no vendor payment for placement. Where a vendor publishes nothing, this page says so rather than estimating. Ranking is by transparency score: pricing transparency 60%, user satisfaction 40%. What a product can do is shown in the condition columns and carries no weight in the number.