{"id":19806,"date":"2026-11-20T23:20:54","date_gmt":"2026-11-20T23:20:54","guid":{"rendered":"https:\/\/shaiksaimeera.in\/blog\/zscaler-adds-dlp-and-malware-scanning-for-claude-what-the-new-genai-coverage-changes\/"},"modified":"2026-11-20T23:20:54","modified_gmt":"2026-11-20T23:20:54","slug":"zscaler-adds-dlp-and-malware-scanning-for-claude-what-the-new-genai-coverage-changes","status":"publish","type":"post","link":"https:\/\/shaiksaimeera.in\/blog\/zscaler-adds-dlp-and-malware-scanning-for-claude-what-the-new-genai-coverage-changes\/","title":{"rendered":"Zscaler Adds DLP and Malware Scanning for Claude: What the New GenAI Coverage Changes"},"content":{"rendered":"<p>Enterprise teams are using AI assistants for more than chat. Claude can become a workspace for drafting, analysis, code, document handling, and other workflows that move sensitive files and messages through a cloud application. That makes the question less about whether an employee can open the service and more about how the organization can see and control the data moving through it.<\/p>\n<p>Zscaler\u2019s Internet &#038; SaaS release summary lists a new feature available on 20 November 2026: DLP and malware-scanning support for Claude. The update lets administrators place Claude under the Gen AI application category, then create separate Data Loss Prevention and Malware Scanning policies for that application tenant.<\/p>\n<p>The change is meaningful because it turns Claude from a broad cloud application entry into a more explicit policy target. The release does not make an organization\u2019s AI usage automatically safe, and it does not replace identity, tenant, or data-classification work. It gives security teams a clearer control surface for a service that may already be part of daily workflows.<\/p>\n<h2>What changed in the Zscaler workflow<\/h2>\n<p>In its <a href=\"https:\/\/help.zscaler.com\/zia\/release-upgrade-summary-2026\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">official 2026 release summary<\/a>, Zscaler says administrators can onboard Claude under <strong>Data Security &gt; DSPM &gt; SaaS Application Tenants &gt; Gen AI<\/strong>. Once the application is configured, DLP policies can be created under <strong>SaaS DLP &gt; Gen AI<\/strong>, while malware policies can be created under <strong>Malware Scanning &gt; Gen AI<\/strong>.<\/p>\n<p>That separation matters. DLP asks whether content contains data the organization does not want sent, stored, or shared through a particular workflow. Malware scanning asks whether a file or object has malicious characteristics. They may inspect the same application path, but they represent different decisions and should be owned, tested, and reviewed as different policy families.<\/p>\n<h2>Before and after for an AI application tenant<\/h2>\n<p>Before an application-specific control exists, an organization may handle Claude through broad web, cloud-app, or Gen AI categories. That can provide visibility, but it may not give administrators enough separation to answer practical questions:<\/p>\n<ul>\n<li>Which policy applies specifically to Claude rather than all AI services?<\/li>\n<li>Which uploaded files were inspected for sensitive content?<\/li>\n<li>Which rule would block or alert on a data-transfer event?<\/li>\n<li>Which files should be treated as potentially malicious?<\/li>\n<\/ul>\n<p>After the update, the organization can create a Claude tenant profile and define DLP and malware controls in the Gen AI area. The result is not a single \u201cAI safe\u201d switch. It is a more legible policy map: the application is identified, the data-protection rules are defined, and malware scanning has its own configuration and response path.<\/p>\n<h2>A concrete example: a document review workflow<\/h2>\n<p>Imagine a finance team using Claude to summarize a set of documents. One file contains internal account information, another contains a public report, and a third arrives from an unknown external source. The workflow needs more than a binary allow-or-block decision.<\/p>\n<p>A DLP policy can identify the sensitive content category the organization cares about and create a record when the file or message matches that policy. A separate malware-scanning policy can inspect the unknown file and produce a different result. The organization\u2019s workflow can then route the sensitive document to a review path while treating the malware result as a file-safety event.<\/p>\n<p>The point is not that the two engines will always produce the same outcome. The point is that the reason for an action remains distinguishable. An analyst reviewing an event should be able to tell whether a file was stopped because it matched a data rule, a malware verdict, or another application-control condition.<\/p>\n<h2>Why tenant-level visibility helps<\/h2>\n<p>Gen AI applications are often used in more than one way. A company may have a centrally managed Claude workspace, individual accounts, API-connected workflows, or separate business units with different data-handling expectations. A tenant-oriented workflow gives administrators a place to distinguish the managed application from an unapproved or unknown use of the same service.<\/p>\n<p>Zscaler\u2019s release summary also lists earlier Claude support in Tenant Profiles, including the ability to provide access to specific workspace IDs. Taken together, the updates point toward a more granular control model: identify the application, narrow the tenant or workspace scope where supported, and apply rules to the traffic or content path that actually matters.<\/p>\n<h2>Trade-offs in adding another policy layer<\/h2>\n<p>More precise policy does not mean less operational work. DLP rules require clear data definitions, test content, owners, and an agreed response when a match occurs. Malware scanning needs a decision about what happens to a suspicious file and how a user is informed. If a team turns on both layers without separating their alerts, analysts may receive duplicate or confusing events.<\/p>\n<p>There is also a tuning trade-off. A strict rule can interrupt legitimate document review, while a broad rule can generate too many events to investigate. The right measurement is not simply how many items were blocked. It is whether the policy distinguishes the events the organization actually needs to understand, while preserving a usable workflow for approved work.<\/p>\n<p>Coverage should also be read carefully. The release describes support for the Claude application in Zscaler\u2019s Gen AI category; it does not establish that every Claude surface, API path, file type, region, or integration has identical behavior. Teams should confirm the supported capabilities and deployment scope in their own tenant before depending on a control for a specific workflow.<\/p>\n<h2>A sensible rollout sequence<\/h2>\n<ol>\n<li>Identify how Claude is used: browser, managed workspace, API workflow, or a combination.<\/li>\n<li>Onboard the correct application tenant and record the workspace scope where the platform supports it.<\/li>\n<li>Start DLP and malware policies in a visibility-oriented mode where possible, then review representative events.<\/li>\n<li>Separate policy owners and response paths so data matches and malware verdicts are not treated as one event type.<\/li>\n<li>Connect enforcement only after testing approved documents, sensitive documents, and untrusted files through the real workflow.<\/li>\n<\/ol>\n<p>Zscaler\u2019s Claude coverage is a product update, but the larger lesson is architectural: AI application security works better when the service, tenant, data policy, and file-safety decision can be seen as distinct parts of the same flow. The memorable question is not \u201cDo we allow Claude?\u201d It is \u201cWhat should happen to each kind of content when Claude becomes part of the work?\u201d<\/p>\n<p><em>This article is a general information summary based on the cited public sources. It is not legal, cybersecurity, or compliance advice.<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Enterprise teams are using AI assistants for more than chat. Claude can become a workspace for drafting, analysis, code, document handling, and other workflows that move sensitive files and messages through a cloud application. That makes the question less about whether an employee can open the service and more about how the organization can see [&hellip;]<\/p>\n","protected":false},"author":0,"featured_media":19807,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[127],"tags":[],"class_list":["post-19806","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-cybersecurity"],"_links":{"self":[{"href":"https:\/\/shaiksaimeera.in\/blog\/wp-json\/wp\/v2\/posts\/19806","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/shaiksaimeera.in\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/shaiksaimeera.in\/blog\/wp-json\/wp\/v2\/types\/post"}],"replies":[{"embeddable":true,"href":"https:\/\/shaiksaimeera.in\/blog\/wp-json\/wp\/v2\/comments?post=19806"}],"version-history":[{"count":0,"href":"https:\/\/shaiksaimeera.in\/blog\/wp-json\/wp\/v2\/posts\/19806\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/shaiksaimeera.in\/blog\/wp-json\/wp\/v2\/media\/19807"}],"wp:attachment":[{"href":"https:\/\/shaiksaimeera.in\/blog\/wp-json\/wp\/v2\/media?parent=19806"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/shaiksaimeera.in\/blog\/wp-json\/wp\/v2\/categories?post=19806"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/shaiksaimeera.in\/blog\/wp-json\/wp\/v2\/tags?post=19806"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}