Teleskope: In-Depth Review & Analysis

Fast Facts

• Headquarters: New York, NY
• Founded: 2022 by Elizabeth (Lizzy) Nammour, formerly of Airbnb security engineering.
• Ownership and funding: private, venture-backed; $32.2M raised, including a $25M Series A led by M13 in November 2025.
• Certifications: SOC 2; ISO 27001.

Teleskope’s Approach

Teleskope discovers and classifies sensitive data, decides what each finding means for the business, and fixes it in one platform. It calls this the Data Reasoning Layer.

Classification runs in two lanes. The first finds sensitive data, using a combination of rules and TelBERT, a task-specific model, and then further validates them using a large language model (LLM). The second assigns each file a document type, such as an offer letter or board memo, and a data subject, such as an employee or customer, assessing the sensitivity of the document itself and in correlation with the data elements inside it.

A risk score then weighs sensitivity, exposure and compliance frameworks, and policies act on it. For example, a public Google Drive link can trigger an email to the owner and, if nobody responds in 14 days, restrict the link to the organization.

Kosmo, the AI agent, routes requests to specialist agents for search, triage, reporting, remediation and policy. Teleskope says only metadata reaches the model provider. We like that Kosmo builds each policy as an editable workflow next to the conversation, so the team sees exactly what will run.

Market Position

Teleskope combines DSPM with cloud and SaaS DLP, retention enforcement and automated labeling for Microsoft Purview. In DSPM it competes with Cyera, Sentra, BigID, Concentric AI, Varonis and Securiti. Teleskope reports winning proofs of concept against Cyera on native remediation and file share coverage, and gaining Varonis customers as Varonis ends self-hosted support on December 31, 2026. In DLP it replaces Nightfall and cloud DLP policies; Ramp moved its Slack and Google Drive DLP into Teleskope. Teleskope is an official Slack and OpenAI partner.

For Microsoft E5 customers it runs alongside Purview rather than replacing it. Accounting firm Aprio’s first Purview scan returned more than 12 million findings, mostly false positives; Teleskope now classifies its documents and applies the labels Purview enforces.

It does not cover USB, print, local transfers or network egress. Buyers who need those should keep an endpoint DLP product such as Purview endpoint DLP, Forcepoint or Netwrix Endpoint Protector. With about 42 staff and no published customer count, Teleskope is far smaller than Cyera, BigID or Varonis.

Deployment is air-gapped, single-tenant SaaS or Teleskope-managed, with the same features in each and agentless API connections. Teleskope sells directly and runs a partner deal registration program.

Use Cases

Finding Sensitive Documents That Pattern Matching Misses

Regex finds an SSN but not an M&A memo with no regulated fields, and it cannot tell a live employee record from test data. Teleskope’s document lane gives each file an AI summary, a document type and a data subject, so a financial forecast is flagged on what it is. A name and email in an employee record count as personally identifiable information (PII); a public figure named in an article does not. For intellectual property, customers supply examples and the engine finds similar files; Teleskope acknowledged this needs company-specific tuning. Teleskope says first findings arrive within about two hours of connecting. Its accuracy figures range from 95% in customer case studies to 99%+ on its product pages, and none is independently tested.

In the demo tenant, a file whose summary described it as synthetic test data still scored 96 for overall risk. We would test how document context feeds the score during a proof of value.

Turning Written Policies Into Enforced Rules

Teleskope said many customers have pages of written data policies they have never put into production. The Policy Document Hub ingests those documents. In the demo, a records retention schedule became ten suggested policies, each linked to its source clause, from retaining governance records permanently to flagging contracts older than five years. Policies are built on a visual canvas from triggers, such as document type, external sharing or MIP labels, and native actions: redact, revoke access including inherited folder access, restrict links, quarantine, tag, delete, or archive to Azure blob storage. Event-driven connectors act on sharing and permission changes as they happen. Policy Simulation, to preview a policy’s effect, is listed as coming soon.

Working Through the Backlog With Kosmo

Teleskope says many customers start with 25,000 or more critical resources. Kosmo’s triage agent groups them by risk: in the demo, 34 critical files with SSNs and bank details already shared externally, with a recommendation to restrict sharing. The remediation agent then removed external users from the selected files in bulk, logging each action. Asked for a public link policy, the policy agent offered three responses, from alert only to immediate restriction, and built the chosen workflow on the canvas. Teleskope reports that its top adopters fully automate 57% or more of exposures.

Stopping Sensitive Data in Chat, Email and AI Tools

Teleskope enforces DLP through APIs and inference hooks, not an agent. It says Slack, Google Chat and Microsoft Teams are monitored in near real time; its Slack app can remove a message, notify the sender or require a business justification. Teleskope also described inline email hooks for Gmail and Exchange that warn or block before sending. For AI tools, Claude Enterprise exposes an inference hook, so prompts and attachments can be blocked before they are sent. ChatGPT and Copilot have no equivalent, so Teleskope monitors them after the event and alerts or deletes. Gemini is listed as coming soon. Buyers should map this against the AI tools they allow.

User Experience

The console centers on a Data Explorer listing every resource with its risk score, sensitivity and data elements. Each resource opens to risk, permissions, findings and activity tabs, and Kosmo sits in a side panel. Teleskope recommends a crawl, walk, run rollout, moving from visibility to automating high-confidence cases, and says every automated action is logged and reversible.

Kosmo builds reports from a prompt. In the demo it produced an AI Enablement Readiness report showing 6,808 files reachable by every internal user, 382 of them critical. We found this a practical answer to the limited reporting Teleskope says customers see elsewhere. Teleskope offers scanning and redaction APIs, alerts to Slack, Jira, Tines and CrowdStrike, and a Model Context Protocol (MCP) server. It says each customer gets a dedicated customer engineer.

Pricing

Pricing is quote-based and depends on the data sources connected, organization size and deployment model. Teleskope says SaaS connectors such as Microsoft 365, Google Drive, Slack and Salesforce are priced per user, with bundles for multiple connectors. Cloud and on-premises data stores are priced by data volume on its SaaS deployment, or as a flat fee per connector on Teleskope-managed and self-hosted deployments. Air-gapped deployments carry an additional support fee. Published case studies describe a proof of value before purchase. Teleskope said it is preparing to publish its pricing.

Strengths and Cautions

Strengths

• Fixes findings natively: redacts, revokes access, restricts links, quarantines and deletes without ticketing.
• Classifies each document’s type and data subject, catching sensitive files regex misses.
• Detects and redacts sensitive data in Slack in under two seconds, Teleskope reports.
• Turns written retention and data handling policies into suggested, enforceable rules.
• Kosmo agent triages risk and builds multi-step remediation policies from plain-language requests.
• Deploys air-gapped, single-tenant SaaS or Teleskope-managed, with identical features.

Cautions

• No endpoint agent or network inspection; USB, print and egress need another product.
• Inline AI blocking covers Claude Enterprise only; ChatGPT and Copilot are monitored after the event.
• Relatively small vendor of about 42 staff, though it has over 25 enterprise customers, including Notion and Polymarket.
• Accuracy claims are unverified, and intellectual property classification needs company-specific tuning.

Who It’s For

We think Teleskope fits security teams that have more sensitive data exposure than people to fix it. The benefit is a shorter path from finding to fix: classification, policy and logged remediation in one platform, with data owners involved only where a policy requires it.

Its customers are weighted toward cloud-first technology companies such as Ramp, Notion and Polymarket, with a newer enterprise base in retail, manufacturing and local government. Teleskope can replace a DSPM tool and cloud or SaaS DLP outright; in Microsoft E5 estates we favor running it alongside Purview.

It is a weaker fit if you need one vendor for endpoint, network and cloud DLP, require inline blocking across every AI tool today, or need a large installed base and independent accuracy testing.