What our analysis found

InterZap, used by 1,000 people, showed no malicious code in Extension Auditor's review. Our code review reported 1 finding (1 medium), led by unauthorized data collection. It comes from a publisher with 20 extensions and was last updated in September 2026.

Key findings

• medium· unauthorized data collection —The CWS listing declares no data collection at all, yet these operator endpoints receive WhatsApp personal data in normal use: ${backend} posts the user's own WhatsApp id, name, profile picture, business email, description and category; ${backend_plugin} posts a contact's digits-only phone plus the account's attribution values; ${backend_utils} uploads the full base64 body of every outgoing non-Opus message for transcoding (Chat handler in); ${backend_utils} and ship funnel and campaign contact rows; and posts {status, contact phone, whatsappID} to the unrelated-domain hook None of it is obfuscated and all of it serves advertised features, but phone numbers, contact identities and message audio are personal data transmitted off-device against a 'none declared' disclosure, and the reseller named on the listing () is not the white-label operator () actually receiving it.

InterZap is a functioning WhatsApp Web CRM matching its listing: the injected page module () is a fixed allowlist of ~100 CRM commands (sendTxt, getMessages, addOrRemoveLabels, userTabs) over the bundled open-source wppconnect wa-js v4.6.1-alpha.0, with no generic function-name passthrough, no remote script (every createElement("script") is React's hoistable-script code, jQuery globalEval, or chrome.runtime.getURL of bundled files), and:44-47 'remote_code' resolving to a version string "1.0.86" plus a DOM-selector — consistent with being dropped in. The MalExt WaSteal IoC does apply to this white-label family (the report's shared platform key ffce211a-7b07-4d91-ba5d-c40bb4034a83 is verbatim at:27), but the two remote-code vectors it attributes to the family (GTM-KMZ9CZK container,) are absent from this build, the alleged "voice interception" is a transcoder applied only to outgoing audio (Chat handler: if e==="sendAudio" &&!base64.startsWith("data:;codecs=opus") then POST to backend-utils), and the daily backup of IA ( keys) and whatsApi tokens is a local Blob download, not an upload. Corroborating suspicion: the IoC listing and CWS-disclosed collection of "none declared" against real flows to the operator (the one finding below); contradicting the ML score: malicious for this publisher, chrome-stats impact and likelihood, 25 reviews describing the actual CRM, and a permhash cluster keyed on the generic [unlimitedStorage, storage, alarms, tabs] set that the same IoC feed appears to have bulk-tagged. Coverage note: the supplied bundle held 9 of 386 files, so I unpacked the full 15 MB CRX for version from local corpus storage and reviewed it.