Google Chrome CVE-2026-106203

Requires the Autofill feature to have been used so that sensitive data is present to be incompletely cleared, and critically requires victim interaction (UI:R): the user must be socially engineered into opening an attacker-crafted HTML page in Chrome for iOS prior to version 155.0.8059.39. There is no authentication requirement (PR:N), but the mandatory user interaction and reliance on social engineering are the primary limiting factors - this cannot be triggered silently or without a user navigating to attacker-controlled content. Exploitation is confined to the affected iOS Chrome builds; updated versions are not affected.

The input CVSS of 8.8 (AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H) is assigned by Chromium's own rating and overstates impact for what the description calls an information-disclosure flaw: 'incomplete cleanup' (CWE-459) in Autofill that lets a remote attacker obtain sensitive information. The I:H and A:H settings in the supplied vector conflict with a pure disclosure outcome - integrity and availability impact are not supported by the description, so real severity is closer to a confidentiality-only issue (my independent vector below rates ~6.5). Exploitation is network-reachable and low-complexity but is NOT zero-click: UI:R plus the explicit 'leveraging social engineering' language means the victim must be lured into visiting a crafted HTML page, which meaningfully lowers practical risk. No EPSS score, KEV listing, or POC was provided; at time of analysis there is no public exploit identified and no evidence of active exploitation, so this reads as a standard browser-patch-cycle item rather than an urgent in-the-wild threat. Priority is routine-to-moderate: patch via normal browser update cadence rather than emergency response.

An attacker crafts a malicious web page and uses social engineering (phishing link, malicious ad, or lure) to convince a Chrome-on-iOS user to open it. Because Autofill does not fully clear previously populated data, the crafted page reads residual sensitive information (e.g., autofill field contents) left over from a prior context. No public exploit code was identified, and exploitation requires the victim to actively visit the attacker-controlled page.