Prompt Security (SentinelOne) Review

Industrial cybersecurity

TL;DR

Last updated: November 3, 2026. Testing track: Documentation Reviewed.

Prompt Security is a GenAI security platform now folded into SentinelOne’s Singularity Platform. It covers employee use of AI tools through a browser extension and lightweight endpoint agent, plus homegrown LLM apps via API, AI agents and MCP servers via a gateway, and red teaming. Pricing and framework mappings are not published, and most performance claims are vendor-supplied.

Ownership: Acquired by SentinelOne (NYSE: S). Announced August 5, 2025; closed September 5, 2025 per SentinelOne’s Form 10-Q, for about $133.6 million in cash plus 1,555,099 Class A shares and 415,109 assumed options (total consideration fair value $160.2 million, preliminary).

Score: 7.2 / 10.

Initial review November 3, 2026.

How This Review Was Conducted

This review is based on Prompt Security’s public documentation, product pages, pricing pages and press coverage as of November 2026. We have not had lab access or a vendor demo.

Scores use the same published rubric and formula as every review on this site. Sources are listed at the end of the review.

Review methodology and evaluation process

Score breakdown

Dimension

Coverage breadth

Detection accuracy

Deployment friction

Policy & control depth

Framework alignment

Pricing transparency

Support & documentation

Weight

20%

20%

15%

15%

10%

10%

10%

Score

9

8

7

8

5

4

6

Notes

Spans browsers, desktop apps, code assistants, homegrown apps, agents and MCP, plus on-premises.

LLM-based contextual detection and prompt injection defense are documented, but no public benchmarks exist.

Extension deploys quickly via MDM, but full coverage adds an endpoint agent and gateway.

Block, redact, alert and coach with role and department scoping, though fine-grained policy documentation is not public.

No published framework mappings found in the sources reviewed.

Quote only, with no marketplace figures found.

Backed by SentinelOne, but product documentation is sparse and support tiers specific to the product are not documented.

Coverage breadth

Weight
20%

Score

9

Notes

Spans browsers, desktop apps, code assistants, homegrown apps, agents and MCP, plus on-premises.

Detection accuracy

Weight
20%

Score

8

Notes

LLM-based contextual detection and prompt injection defense are documented, but no public benchmarks exist.

Deployment friction

Weight
15%

Score

7

Notes

Extension deploys quickly via MDM, but full coverage adds an endpoint agent and gateway.

Policy & control depth

Weight
15%

Score

8

Notes

Block, redact, alert and coach with role and department scoping, though fine-grained policy documentation is not public.

Framework alignment

Weight
10%

Score

5

Notes

No published framework mappings found in the sources reviewed.

Pricing transparency

Weight
10%

Score

4

Notes

Quote only, with no marketplace figures found.

Support & documentation

Weight
10%

Score

6

Notes

Backed by SentinelOne, but product documentation is sparse and support tiers specific to the product are not documented.

What it does well

One product covers employees, apps and agents.

Covers employee AI use, code assistants, homegrown LLM apps, agents and MCP in one product line, which few workforce-focused tools match (AppSecSanta review; SiliconANGLE, March 2026).

Offers redaction, coaching and role-based policies.

Offers redaction and coaching, not only blocking, with role and department based policies (NAND Research; AppSecSanta).

Red teaming added for homegrown apps.

Red teaming and runtime guardrails for homegrown apps were added as SentinelOne offerings in March 2026 (Security Boulevard, RSAC 2026).

Announced on-premises, air-gapped deployment option.

An on-premises, air-gapped deployment option was announced for data sovereignty requirements (SiliconANGLE, March 2026). Backed by a public company, with the acquisition closed and documented in SEC filings (SentinelOne Form 10-Q).

Where it falls short

Pricing not published; quoted via SentinelOne.

Pricing is not published and must be quoted through SentinelOne (AppSecSanta).

No published framework mappings found.

No published mappings to NIST AI RMF, OWASP, ISO 42001 or the EU AI Act were found (not documented).

Detection and latency claims are vendor-supplied.

No independent detection benchmarks; the 15,000+ AI tools and latency claims are vendor-supplied (AppSecSanta; Techzine). Full coverage depends on deploying browser extensions and endpoint agents, adding endpoint management work (NAND Research). Product direction now follows SentinelOne's platform roadmap, and public documentation has moved behind sentinelone.com (AppSecSanta).

Best fit

Security teams, especially existing SentinelOne Singularity customers, that want one vendor for employee AI usage control plus guardrails for homegrown LLM apps, AI coding assistants, agents and MCP servers, and who value redaction over hard blocking. Regulated or air-gapped environments may also value the on-premises option.

Poor fit

Buyers who need transparent pricing, published framework mappings, or board-ready AI governance reporting, and those who want agentless deployment. Alternatives include Microsoft Purview for Microsoft-centric estates, Harmonic or Nudge for workforce AI usage with lighter governance overhead, Lakera for application-layer prompt defense with public benchmarks, and Island where an enterprise browser is acceptable.

Pricing transparency

Not published. Contact SentinelOne; no public AWS or Azure marketplace price found.

Alternatives

See how it compares in our Best AI DLP Tools 2026 ranking. The poor-fit section above names alternatives for buyers it does not suit, and our methodology explains how every product is scored.

What We Would Test in the Lab

If Prompt Security grants lab access, we would run the following scenarios and upgrade this review to Lab Tested.

Sensitive data detection.

The standard sensitive-data set (PII, PHI, payment data, secrets, source code) sent through the product's protected AI tools.

Shadow AI discovery.

Inventory accuracy against a known set of sanctioned and unsanctioned AI tools.

Policy enforcement.

Block, warn, redact and allow behaviors against a configured policy.

Audit logging.

What is logged, what is not, and retention behavior.

SSO integration.

Microsoft Entra ID and Okta where supported.

Deployment.

Time from access to enforced policy in a representative environment.

Latency.

Added latency on standard prompt sizes.

Feature coverage at a glance

Based on public documentation. “Not documented” means we could not find it in public material, not that the product lacks it.

CapabilityPrompt Security
Inline prompt DLPYes. Inspects prompts in real time and redacts or blocks sensitive data before it reaches the AI tool.
Shadow AI discoveryYes. Discovers sanctioned and unsanctioned AI tools via extension and agent; vendor claims 15,000+ AI sites.
Blocking unapproved AI toolsYes. Policies can block high-risk interactions and tools.
File upload inspectionYes. Third-party review states the extension monitors prompts, file uploads and responses.
Secure AI workspace (sanctioned chat)Not documented. No vendor-provided sanctioned chat interface documented.
Built-in AI vendor risk scoringNot documented. No built-in AI vendor risk ratings documented in sources reviewed.
Microsoft 365 connected AI app review and controlNot documented. No Microsoft 365 OAuth app review or revoke documented.
Google Workspace connected AI app review and controlNot documented. No Google Workspace connected-app governance documented.
Governance scorecard, attestation and board reportingPartial. Searchable audit trail of AI interactions; board-level scorecards or attestation not documented.
AI agent and MCP securityYes. Prompt AI Agent Security discovers MCP servers used by agents and enforces guardrails through an MCP gateway.
Prompt injection defenseYes. Detects prompt injection and jailbreaks for homegrown apps and agents.
AI red teamingYes. Prompt AI Red Teaming (announced March 2026) simulates prompt injection, jailbreak, privilege escalation and data poisoning; open-source ps-fuzz also available.
SIEM integrationPartial. Being integrated into SentinelOne Singularity and its AI SIEM; third-party SIEM export not documented in sources reviewed.

Recent changes

  • 2025-08-05: SentinelOne announced its agreement to acquire Prompt Security.
  • 2025-09-05: Acquisition closed (SentinelOne Form 10-Q).
  • 2025: SentinelOne began integrating Prompt Security into the Singularity Platform, launching Prompt Security for Employees among four Prompt-branded products (Techzine).
  • 2026-03-23: At RSAC 2026, SentinelOne announced Prompt AI Agent Security, Prompt AI Red Teaming and Prompt Security On-Premises.

Sources

Deployment and coverage

Browser extension, a lightweight endpoint agent, API integration for homegrown apps, and an MCP gateway; delivered as SaaS within SentinelOne Singularity, with an on-premises option announced March 2026. Third-party review says the extension deploys in minutes; NAND Research notes agents and extensions add endpoint management burden.

Employee use of web AI tools (vendor claims 15,000+ AI sites), desktop AI applications, AI code assistants, homegrown LLM applications via API, AI agents and MCP servers via gateway, plus on-premises and air-gapped deployments.

FAQ

Who owns Prompt Security?

SentinelOne. The acquisition was announced August 5, 2025 and closed September 5, 2025, according to SentinelOne’s SEC filings.

Through a browser extension and lightweight endpoint agent for employees, an API for homegrown LLM apps, and a gateway for MCP servers; an on-premises version was announced in March 2026.

Yes. Prompt AI Agent Security identifies MCP servers used by agents and applies guardrails, including blocking agents from sending business data to third parties.

SentinelOne does not publish pricing; it is quoted per customer.