Protecto

• Microsoft and AWS already sell the sensitive-data detection Protecto measures itself against, so either could extend that detection into per-user decisions at inference and remove the third-party line item.
• Every named deployment reaches the public through Protecto's own case studies, so a procurement team that requires an independent reference could rule the company out before a technical evaluation begins.
• Protecto has announced no funding since November 2023, so a rival raising at agent-era valuations could outspend it on enterprise sales before its named accounts turn into a referenceable base.
• Protecto backs its detection quality with a benchmark study Protecto's own page summarizes and gates behind a form, so an independent evaluation reporting weaker detection than the 99.9% figure would undercut the core claim.
• Protecto publishes no price and its pricing address returns a not-found page, so a developer evaluating the API has to request a trial or book a demo before learning what it costs, which slows the self-serve motion the documentation targets.
• A TechCrunch teardown sets the CTO's stated 18 years at Apple against a LinkedIn record of seven, so a buyer's diligence could find the privacy-engineering depth thinner than the about page states.

Protecto treats the data an AI system assembles at inference as the thing to control. The homepage addresses the enterprise team running LLMs, agents and MCP pipelines, and describes a platform that detects sensitive information, masks it in real time, enforces who sees what, and logs everything.

The problem statement is specific about where older controls stop. The company's access-control page argues that controls built for files, folders, apps and databases lose visibility once content enters a prompt or payload, and that user access is not task access.

What the reviewed record does not carry is an outside measure of that pain. The independent sources here cover a funding round, a product launch and a pitch deck, and none of them quantifies how often sensitive data reaches a model it should not reach, so the scale of the problem stays the company's own account. [s1, s5, s10, s11, s13]

Protecto is an API a developer calls where data enters or leaves a system. The documentation describes a platform that identifies, tokenizes, masks and safely handles sensitive data across applications, pipelines and GenAI workflows, and the homepage counts more than 200 sensitive data types.

Tokenization is the technical bet, and the architecture page states the mechanism. Sensitive values become structured tokens that preserve type, position and referential relationships, the same input always produces the same token within a namespace, and a vault holds the mapping back to original values under access policy.

Access decisions run during inference rather than at the source. Context-Based Access Control weighs who is asking, what they are asking and which policy applies, returning only the parts a role, task and context should see, with masking and unmasking tied to directories such as Active Directory and Okta.

Verification stops at Protecto's own account. The homepage advertises 99.9% detection accuracy and states that the product was independently verified to outperform AWS Comprehend and Microsoft Presidio, while the page describing that third-party study keeps the study behind a download form. [s3, s18, s5, s1, s9]

Protecto names its own rivals, and the list spans four kinds of product. Its comparisons page publishes head-to-head pages against SkyFlow, Microsoft Presidio, AWS Comprehend, John Snow Labs and Protegrity, describing Presidio as Microsoft's pattern-based approach and Comprehend as AWS's NLP-focused solution.

Two of those five carry the names of products from Microsoft and AWS. That places the detection half of the product beside capability the same vendors sell, and Protecto's published answer is a third-party benchmark study its own page summarizes and gates behind a form.

An aggregator ranks the field differently again. Tracxn lists Enkrypt AI, Protect AI and SSI as top competitors and places Protecto 11th among 306 active competitors, so which contest a buyer sees depends on which list they read. One customer quote on Protecto's own case study puts the choice against John Snow Labs. [s8, s9, s15, s6]

Protecto's traction record now names customers, and Protecto publishes all of it. Its case studies name Inovalon, a health analytics company that deployed the product entirely within its own cloud environment and tested more than 1,000 documents in under a week, and Automation Anywhere, whose secure-agent rollout the case study puts at 5,000-plus enterprise customers.

Distribution moved in March 2026. Protecto Vault became available on Google Cloud Marketplace, procurable against existing Google Cloud committed spend, and the company's release quotes Dai Vu of Google Cloud on the listing.

The independent record is thinner. Help Net Security covered the Vault launch in December 2025 and quotes Venkat Kam, chief executive of Revinci AI, and the 2023 funding coverage named Brookfield Renewables and Nokia as initial customers. No company-reported revenue and no third-party count of Protecto's own customers appear, the outside revenue signal is an aggregator estimate of under 10 crore rupees as of March 2025, and Protecto's product pages claim data protection for more than 3,000 enterprises while the automation case study counts more than 5,000 downstream customer tenants on one platform, so neither figure is a third-party count of Protecto's own customers.

Buying starts with a conversation. Protecto publishes no price, its pricing address returns a not-found page and the published sitemap lists no pricing page, so the site routes a prospect to a trial request form answered within one business day. [s6, s7, s14, s10, s13, s16, s5, s15, s19]

Protecto's founders pair product and data engineering rather than a security pedigree. Founder and CEO Amar Kanagaraj scaled a previous startup as its chief marketing officer, led product management for Microsoft Search and AI, and started as a developer at Sun Microsystems.

One founder claim is disputed in public. The about page credits co-founder and CTO Baskaran Alagarsamy with more than 18 years at Apple leading privacy engineering, while a TechCrunch teardown of the seed deck reports seven years as a manager at Apple India on his LinkedIn and asks about the gap, so the privacy-engineering depth the about page claims is contested in the public record.

The advisory bench is investor-led. Girish Mathrubootham, founding partner at Together.Fund and founder of Freshworks, and Lakshmi Shankar, a general partner there and formerly a vice president at Google Search and Gemini AI, are the named advisors, and the about page states Google chose the company for a 2025 accelerator from more than 1,600 applications. [s2, s11]

Protecto states the attestation set an enterprise data buyer expects to see. The homepage lists SOC 2 Type II, ISO 27001, HIPAA with a business associate agreement, GDPR, DPDP and CPRA, with audit logs covering every scan, mask and unmask event.

No trust portal and no downloadable report appear in the reviewed record. A security review would request the reports and the data-handling terms from the company directly.

Deployment options carry the regulated buyer's requirements. Protecto runs in SaaS, a private virtual private cloud, or fully on-premises environments, and the company publishes a support agreement that categorizes issues by criticality and reports weekly on the critical and high ones. [s1, s16, s20]

Add analyzed competitors to compare them side by side with Protecto.

Protecto sells to the enterprise team that has to put regulated data through AI. The documentation addresses engineering teams identifying, tokenizing and masking sensitive data across applications, pipelines and GenAI workflows, and the site groups its solutions by healthcare, financial services, B2B SaaS, government and insurance.

The clearest named deployment sits at the regulated end of that range. Inovalon, a health analytics company, deployed the product entirely within its own cloud environment for HIPAA compliance, and the offer carries a business associate agreement with on-premises and air-gapped deployment for regulated industries.

Who signs is less visible than who uses. A prospect reaches a trial request form rather than a published price, and The launch coverage quotes a startup chief executive quoted in the launch coverage and an unnamed platform-engineering vice president in Protecto's own case study, so the budget holder stays off the record.

Protecto's core capability is finding sensitive data in messy text and replacing it with tokens a model can still reason over. The architecture page describes AI-driven classification across structured and unstructured content, and semantic tokens that preserve the type, position and referential relationships of the original data so a document's entities stay linked.

The second capability decides, per request, what a model may return. Context-Based Access Control weighs identity, prompt intent, data sensitivity and policy in real time during inference, and access policies govern the route back to original values so only authorized workflows restore them.

Engineering output is documented and dated. Vault release notes run from June 2024 to version 8.0.6 in June 2026 and record new healthcare and regional entity types plus multilingual API support, so a buyer can see what changed and when.

Quality is asserted rather than shown. The homepage advertises 99.9% detection accuracy and states that the product was independently verified to outperform AWS Comprehend and Microsoft Presidio, and the page describing that third-party study keeps the study behind a download form.

Protecto's go-to-market runs through the product and, since March 2026, through a cloud provider's storefront. Protecto Vault became available on Google Cloud Marketplace, runs entirely within the customer's Google Cloud environment with no external data movement, and can be procured against existing Google Cloud committed spend.

Named proof exists, and Protecto publishes all of it. Its case studies name Inovalon, which tested more than 1,000 documents in under a week and deployed within its own cloud environment, and Automation Anywhere, whose secure-agent rollout the case study puts at 5,000-plus enterprise customers. The about page's image alt attributes name Inovalon, Automation Anywhere and Bank of Muscat, and the homepage's served structured data describes a video in which Steve Shah, SVP of Products at Automation Anywhere, discusses Protecto's part in securing enterprise data.

Outside those pages the record thins. Help Net Security covered the December 2025 Vault launch and quotes Venkat Kam, chief executive of Revinci AI, and the November 2023 funding coverage named Brookfield Renewables and Nokia as initial customers. No company-reported revenue and no third-party count of Protecto's own customers appear, the outside revenue signal is an aggregator estimate of under 10 crore rupees as of March 2025. Protecto's product pages claim data protection for more than 3,000 enterprises while the automation case study counts more than 5,000 downstream customer tenants on one platform, so neither figure is a third-party count of Protecto's own customers.

Protecto publishes no price. The pricing address returns a not-found page and the published sitemap lists no pricing page, so nothing on the site lets a buyer size the spend before contacting the company.

A conversation replaces the price page. The site routes a prospect to a free-trial request form answered within one business day, which puts a person between the developer and any spend.

One pricing signal survives outside the site. Help Net Security's launch coverage describes Protecto Vault as carrying a low-cost, pay-as-you-go model, which states a direction rather than a rate.

Protecto is delivered as an API the customer wires into its own systems. A developer calls the service where data enters or leaves, and the product handles detection, masking and optional unmasking under strict authorization, so the customer's team operates it and owns the outcomes.

Deployment stretches to the regulated end. The product runs in SaaS, a private virtual private cloud, or fully on-premises environments, and the Google Cloud listing runs inside the customer's own cloud with no external data movement.

Support carries a published commitment. Protecto publishes a service-level agreement for product support that categorizes issues by criticality and promises weekly status reports on critical and high issues.

Protecto states the certifications an enterprise data buyer asks for. The homepage lists SOC 2 Type II, ISO 27001, HIPAA with a business associate agreement, GDPR, DPDP and CPRA, with audit logs covering every scan, mask and unmask event.

No trust portal and no downloadable report appear in the reviewed record. A buyer confirms the attestations by asking Protecto for the reports rather than by reading them.

The controls themselves are ordinary for this market. A funded competitor can obtain the same certifications through normal enterprise preparation, so they ease a procurement review without blocking a substitute.

Protecto positions one layer between enterprise data and every AI system rather than a point tool. The homepage frames it as the plane controlling the enterprise data that flows through identity systems and gateways, and the product line now carries five named modules, the Data Privacy Vault, CBAC, DeepSight, GPTGuard and High-volume Masking.

Reach comes through integrations rather than a builder community. Trade-press launch coverage names n8n, LangGraph, Zapier and MCP servers as the ecosystems Protecto Vault integrates with, and the product surface is REST APIs and a Python SDK rather than a partner marketplace of its own.

The layer it occupies is one larger vendors also sell into. Protecto's own comparison pages measure the product against Microsoft Presidio and AWS Comprehend, so the detection half of the platform stands beside offerings from the clouds its buyers already run.

Protecto's leadership pairs a product-and-growth founder with a data-engineering co-founder. Amar Kanagaraj scaled a previous startup as chief marketing officer and led product management for Microsoft Search and AI, which TechCrunch puts at close to eight years at Microsoft.

The chief technology officer's record is disputed in public. The about page credits Baskaran Alagarsamy with more than 18 years at Apple leading privacy engineering, and a TechCrunch teardown of the seed deck reports seven years as a manager at Apple India on his LinkedIn, so the depth the about page claims is contested in the public record.

The backing is operator-led and small. A $4 million seed from Together Fund with Better Capital, FortyTwo VC, Arali Ventures and Speciale Invest followed 2021 angel capital from Google and Microsoft executives, for $5 million in total, and an aggregator puts the company at 40 employees as of June 2026, so an enterprise motion runs on seed-stage capital.