TL;DR: Cyberhaven and Code42 Incydr both focus on insider-risk detection through endpoint visibility and data-lineage-style evidence, but the article argues that neither is built to remediate exposure inline, leaving SaaS, browser, and AI-agent leak paths undercovered, according to Strac. The broader issue is that visibility-first DLP still leaves organisations with proof after exposure rather than prevention at the point of data movement.
At a glance
What this is: This comparison says Cyberhaven and Code42 Incydr are strong on insider-risk visibility, but weak on inline remediation and broad coverage across SaaS, browser, GenAI, and AI-agent leak paths.
Why it matters: It matters because IAM, PAM, NHI, and data-security teams need controls that stop sensitive data movement at the point of exposure, not only evidence after the fact.
Context
Cyberhaven and Code42 Incydr sit in the visibility layer of insider-risk control. They help teams observe how data moves, who touched it, and where exfiltration may have occurred, but that is not the same as stopping the exposure itself.
The governance gap is that many organisations still separate detection from enforcement, especially across SaaS, browsers, GenAI prompts, and AI-agent traffic. When those channels carry sensitive data, endpoint-centric monitoring alone leaves the control point too late in the workflow.
Key questions
Q: What breaks when DLP only detects data after it has already moved?
A: When DLP is detection only, the security team learns about exposure after the data has already been copied, shared, or transformed in another system. That creates blind spots across collaboration tools and AI workflows. Effective DLP needs enforcement at the point of risk, with controls that can stop or remediate exposure in real time.
Q: Why do browser prompts and AI-agent traffic evade endpoint-led DLP models?
A: Because the risky exchange often happens inside web sessions and connector calls rather than through a classic file-transfer event. Endpoint visibility can still help, but it does not automatically inspect or modify the content in the browser or the agent workflow. The result is a control gap between seeing the action and enforcing on the payload.
Q: What signs show that a DLP program is only producing evidence after exposure?
A: If the main outputs are alerts, investigations, and user-risk scores while sensitive content still reaches SaaS, chat, or AI tools, the program is evidence-first. Another sign is that containment depends on manual response after the transfer rather than automated redaction or blocking before transmission.
Q: Should teams prioritise inline remediation over more visibility in insider-risk tools?
A: If the business goal is to prevent sensitive data from leaving approved boundaries, yes. Visibility is still useful for detection and forensics, but it does not reduce exposure on its own. Teams should prioritise inline remediation whenever regulated, confidential, or secret data can move through browser, SaaS, or AI workflows.
Technical breakdown
Data lineage versus inline remediation
Data lineage tools trace a file or payload across endpoints and applications so investigators can reconstruct origin, movement, and transformation. That is useful for forensics and insider-risk scoring, but it is not the same as enforcing policy on the data before it leaves the environment. Inline remediation sits closer to the transfer point and can redact, mask, block, or quarantine content before transmission. The architectural difference is simple: lineage answers what happened, while remediation changes whether exposure happens at all.
Practical implication: Treat lineage as evidence generation, not exposure prevention.
Why endpoint-only coverage misses GenAI and MCP traffic
Endpoint agents are effective when the risky action happens on the device, but modern leak paths increasingly span browser-based GenAI prompts and AI-agent tool calls through MCP connectors. In those cases, sensitive data can move through web sessions, SaaS integrations, or agent workflows without looking like a traditional file exfiltration event. That means the control surface must extend beyond the laptop and into application, browser, and connector layers if the organisation wants to intercept the data before it is consumed downstream.
Practical implication: Map policy enforcement to browser, SaaS, and agent connectors, not only to the endpoint.
Detect-and-alert leaves the last mile unresolved
Detect-and-alert programs are built to create visibility, prioritise response, and support investigations. They are not built to remove sensitive content from the object being sent, whether that is an email, a SaaS message, or a prompt submitted to an AI service. Once the model is tuned to alert after risky movement, response teams still have to chase containment, revoke access, or clean up exposure after the transfer. That is a governance gap, not merely a tooling preference.
Practical implication: Align DLP outcomes to exposure prevention, not just incident documentation.
Threat narrative
Attacker objective: The objective is to move sensitive data out of approved control boundaries while avoiding enforcement that would block, redact, or quarantine it.
- Entry occurs when sensitive data is handled on an endpoint, in a SaaS workflow, or inside a browser-based AI prompt rather than in a tightly controlled repository.
- Credentialed or authorised access is then used to move content into external services, cloud apps, or AI-agent tool calls without inline inspection stopping the transfer.
- Impact is exposure of regulated or confidential data, followed by forensic reconstruction after the fact instead of prevention at the point of movement.
Breaches seen in the wild
- Cyberhaven Chrome extension breach 2024: A phished OAuth consent gave attackers Chrome Web Store publishing rights and a malicious Cyberhaven extension update reached about 400,000 users.
Read and download The State of NHI & AI Agent Breach Report 2026, covering 200+ breaches impacting Non-Human Identities including AI Agents.
NHI Mgmt Group analysis
Visibility-first DLP creates a remediation gap, not a complete insider-risk program. Endpoint telemetry and data lineage are valuable, but they stop at observation unless enforcement exists at the point of movement. In 2026, the practical distinction is between proving a leak and preventing one, and teams should judge tools by that boundary.
Browser and AI-agent data paths are now first-class governance problems. GenAI prompts and MCP-connected agent traffic are not side channels anymore, they are routine exposure paths for sensitive data. That means insider-risk policy has to cover the browser, SaaS applications, and agent connectors together, or data will simply route around the endpoint control.
The named concept here is remediation-first data security. That is the point where detection, policy, and enforcement converge on the data itself rather than on post-incident evidence. The more sensitive workflows move into SaaS and AI interfaces, the more remediation-first design becomes the deciding control posture for practitioners.
Human-risk suites do not remove the need for inline containment. Folding insider-risk tooling into a broader human-risk platform may help with workflow consolidation, but it does not solve the last mile of exposure. Security teams should separate investigative value from control value and insist on both.
From our research library:
- The average time to mitigate a leaked secret is 36 hours, highlighting the operational burden of manual remediation processes, according to the 2024 State of Secrets Management Survey.
- The average estimated time to remediate a leaked secret is 27 days, despite 75% of organisations expressing strong confidence in their secrets management capabilities, according to the State of Secrets in AppSec.
- Read next: MCP Security Guide
What this signals
Remediation-first data security is becoming the more useful control pattern. When data moves through SaaS, browsers, and AI connectors, the important question is no longer only whether the movement can be seen. The better programme design is the one that can change the payload before it crosses the boundary, because post-event visibility does not reduce the initial exposure.
Insider-risk programs need broader coverage than endpoint telemetry alone. The control plane now extends into browser sessions, SaaS storage, and AI-agent connectors, which means DLP and DSPM decisions should be evaluated together. Teams that keep these surfaces separate will continue to find gaps where the user experience changes faster than the policy stack.
Browser and MCP traffic are where leakage assumptions are breaking first. If a program still treats GenAI prompts and agent tool calls as edge cases, it will undercount its highest-risk paths. Security teams should assume that these channels are becoming normal data movers, not niche exceptions.
For practitioners
- Define the enforcement boundary for data movement Document exactly which surfaces must be able to redact, mask, block, or quarantine sensitive content before transmission, including SaaS, browser, endpoint, and AI-agent traffic.
- Separate forensic visibility from prevention requirements Classify endpoint lineage and user-risk scoring as investigative controls, then verify which workflows still need inline remediation to stop exposure at the source.
- Review GenAI and MCP exposure paths Inventory browser prompts, SaaS copilots, and MCP connector flows to see where sensitive data bypasses existing DLP inspection and response rules.
- Test remediation before alert workflows Run tabletop and live tests that prove the platform can alter or stop the content itself, not merely generate an alert after the transfer completes.
- Re-score insider-risk tooling by control objective Use a simple question in procurement and renewal reviews: does the tool change the data before it leaves, or only help investigate the loss afterward?
Key takeaways
- Cyberhaven and Code42 Incydr are positioned as visibility-heavy insider-risk tools, but their comparison exposes a gap between detecting movement and stopping exposure.
- The article’s central concern is that SaaS, browser, GenAI, and MCP-driven AI-agent paths sit outside the control model that endpoint-first DLP products were built around.
- Practitioners should evaluate whether a data-security program can modify the payload before transmission, because investigation after the fact is not the same as prevention.
Standards & Framework Alignment
This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.
OWASP Non-Human Identity Top 10, MITRE ATT&CK and OWASP Agentic AI Top 10 address the attack and risk surface, while NIST CSF 2.0 and NIST SP 800-53 Rev 5 set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| OWASP Non-Human Identity Top 10 | NHI-02 — Secret Leakage | The article centers on sensitive data and secret leakage through browser, SaaS, and AI-agent paths. |
| NHI-10 — Human Use of NHI | Browser and MCP-connected agent workflows create human-initiated NHI exposure paths. | |
| Recommendation — Apply NHI-02 controls to detect and stop sensitive data leaving approved workflows. Restrict human-initiated use of NHIs in browser and agent flows to approved, inspectable channels. | ||
| MITRE ATT&CK | TA0009;TA0010 — Collection; Exfiltration | The comparison is about collecting sensitive data and moving it out through external channels. |
| Recommendation — Map insider-risk telemetry to collection and exfiltration stages, then test for prevention gaps. | ||
| NIST CSF 2.0 | PR.DS-01 — Data-at-Rest Is Protected | The article highlights exposure across SaaS data at rest and moving data, both governed by data protection. |
| PR.DS-10 — Data in Transit Is Protected | Inline redaction and blocking are about protecting data while it moves through browser and AI channels. | |
| Recommendation — Extend PR.DS-01 controls across SaaS and collaboration surfaces where sensitive data sits at rest. Apply PR.DS-10 protections to traffic that carries sensitive data through browsers and connectors. | ||
| NIST SP 800-53 Rev 5 | SC-7 — Boundary Protection | The core issue is whether controls stop data before it crosses the boundary into external services. |
| Recommendation — Use SC-7 to enforce policy at boundary-crossing points where sensitive content leaves the enterprise. | ||
| OWASP Agentic AI Top 10 | ASI03 — Identity & Privilege Abuse | MCP-connected agents can move sensitive data through delegated access paths if privileges are not constrained. |
| Recommendation — Constrain delegated agent privileges and inspect tool calls for unintended data movement. | ||
Key terms
- Data Lineage: The record of how data moves across systems, applications, and workflows. In security operations, lineage shows where sensitive data propagates, which identities touch it, and how a compromise could spread across connected environments.
- Inline remediation: Inline remediation is the practice of presenting security guidance directly in the developer environment where code is written. It reduces context-switching and can speed up fixes, but it only improves governance when the guidance is accurate, explainable, and consistently adopted by engineering teams.
- Browser-Side Data Exposure: The unintentional disclosure of sensitive information through code that executes in a browser session. This can happen through malicious scripts, misconfigured integrations, or legitimate tags that collect and forward more data than the business intended.
- MCP Traffic: MCP traffic is the data exchanged between an AI agent and the tools or services it uses through the Model Context Protocol. It includes requests, responses, context, credentials, and tool outputs. Security teams monitor it to understand what an agent can access, what it sends, and whether actions are authorized and traceable.
Deepen your knowledge
NHI Mgmt Group covers identity security, NHI governance, and agentic AI through independent research, practitioner guides, and the NHI Foundation Level course. It is designed for security practitioners who need to connect identity governance to workload, service, and agentic access patterns.
Published by the NHIMG editorial team on August 20, 2026.
Updated on October 12, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org