New CSS Attacks Can Break Webmail Defenses to Steal Passwords and Tokens

What happened

New CSS attacks have been uncovered that allow malicious content inside emails to escape its intended message boundaries and manipulate the webmail interface. These attacks affect major providers including Outlook, Gmail, Fastmail, Proton Mail, Yahoo Mail, and AOL Mail. The vulnerabilities enable attackers to capture passwords, hijack tokens, take control of third-party accounts, and interfere with trusted interface elements and AI tools that process email content. The research exposing these methods was conducted by PortSwigger researcher Gareth.

The risk

The core risk here is that email, typically isolated into safe containers within webmail, can be exploited to cross those boundaries and break the expected security model of email clients. Attackers can use CSS alone to extract sensitive information visible in the interface or trigger unintended actions without user consent. This is particularly concerning given the widespread use of webmail for sensitive communications and access to various linked services. AI-driven email tools that parse user messages are also targets, potentially exposing them to manipulation or breach.

Why it matters

Webmail providers have long relied on containment boundaries to prevent email content from interfering with the user interface or leaking critical data. These new CSS-based techniques undermine that defense, forcing providers to revisit isolation strategies and potentially redesign rendering engines. For operators, this raises the cost and complexity of securing email platforms and maintaining user trust. Businesses and users relying on webmail for sensitive communications face increased exposure to credential theft and account takeovers. The involvement of AI tools in the attack chain heightens risk by expanding attack surfaces beyond traditional email rendering.

Who should pay attention

Webmail providers, security teams, and developers building email clients or integrations should prioritize this threat. Companies using webmail for business-critical workflows or identity management must reassess their risk exposure. AI developers working on email parsing and automation tools should also run threat models against these CSS vulnerabilities. Security operations and incident response teams should monitor for signs of account compromise tied to webmail interfaces.

What to watch next

Look for responses from major email providers addressing these CSS attack vectors through patched rendering, stricter content isolation, or revamped UI architectures. Updates in browser security models that reinforce iframe and CSS sandboxing could shape the landscape. The pace and scope of mitigation will pressure email platform operators to tighten defenses or risk losing user trust. Follow research tracking variations or exploit attempts targeting AI-driven email processing tools as part of this evolving threat.

AI Quick Briefs Editorial Desk