Detailed Analysis
A user attempting to create and access a Claude account reports a persistent authentication failure tied specifically to the email verification flow. The core issue centers on a mismatch between the verification method expected and the one delivered: the system prompts the user for a numeric verification code, but the email received contains only a link that resolves to a broken web page. The user notes this problem has recurred over a period of several months, affects multiple devices including both a smartphone and a Linux desktop, and appears to be specifically associated with the use of a Proton Mail address.
The significance of this report lies in several compounding factors. First, the persistence of the bug over months suggests this is not an isolated or transient infrastructure hiccup but a reproducible failure state in Anthropic's onboarding pipeline. Second, the cross-device reproducibility rules out client-side or browser-specific explanations, pointing instead to a server-side or email-routing-level defect. Third, the involvement of Proton Mail is a meaningful data point: Proton Mail is an encrypted, privacy-focused email provider that uses non-standard routing and security infrastructure, which can trigger false positives in spam filters, alter link-click tracking behavior, or interfere with tokenized authentication URLs. The broken landing page is consistent with an expired, malformed, or improperly encoded verification token that fails to resolve correctly once the link is accessed.
From a broader perspective, this issue reflects a known tension in identity verification design for AI platforms. Many services default to "magic link" authentication as the primary flow but implement a secondary numeric-code fallback that is either inconsistently surfaced or insufficiently documented. When the primary flow fails — particularly for privacy-conscious users who favor services like Proton Mail — the absence of a clearly available alternative path creates a hard lock on account access. This is a significant accessibility and inclusivity gap, as Proton Mail has tens of millions of users, many of whom are precisely the technically sophisticated demographic likely to engage with AI tools like Claude.
The report also touches on a recurring challenge in onboarding trust and user retention. A new user encountering a broken verification loop before ever successfully using the product faces a disproportionately negative first impression. For Anthropic, which competes in a crowded large language model market, failed onboarding represents not just a technical defect but a reputational and conversion risk. The fact that this user has attempted resolution twice over several months without success indicates that whatever fixes may have been deployed were either incomplete or did not address the Proton Mail-specific routing edge case. A robust resolution would likely require either direct QA testing against Proton Mail's delivery and link-handling behavior or the implementation of a more prominent, always-available fallback verification method that does not depend on successful link traversal.
Read original article →