Detailed Analysis
A documented authentication failure affecting Claude's macOS platform reveals a structural gap in Anthropic's sign-in architecture: accounts created via "Sign in with Apple" on iOS have no corresponding login pathway on macOS, leaving affected users — including paying Max plan subscribers — unable to access the service on desktop entirely. The mechanism of the failure is precise and reproducible. When an Apple Sign-In user attempts to authenticate on macOS through the email-based magic link or six-digit code flow, the system successfully verifies their email but then routes them to the onboarding screen for new account creation rather than recognizing the existing account. The "Sign in with Apple" button, which serves as the original authentication provider for these users, is simply absent from the macOS login interface in both the native desktop application and all tested browsers. This creates a logical dead end: the user cannot proceed through the Apple Sign-In path because it doesn't exist on the platform, and proceeding through the email path risks generating a duplicate account under a different authentication provider — potentially severing access to an active paid subscription permanently.
The scope of the issue extends beyond a single user report. Two open GitHub issues on Anthropic's claude-code repository — filed in March and April of 2026 respectively — document the identical redirect loop behavior with no posted resolution, suggesting the bug has persisted across multiple months and affects a meaningful segment of users who chose Apple Sign-In during account creation. The platform asymmetry at the root of the problem reflects an architectural decision, likely deliberate given Apple's Sign-In with Apple framework requirements for iOS apps, that was apparently never extended to macOS equivalents. Whether that reflects a technical limitation of Apple's cross-platform authentication APIs or an implementation gap on Anthropic's side is unclear from available information, but the practical result is that iOS-originated Apple accounts exist in a class that macOS infrastructure does not recognize as a valid returning user state.
The support response timeline compounds the technical failure with a customer service breakdown of notable duration. A detailed ticket submitted on May 3rd received only automated bot responses through an AI agent called Fin for eight days before a ticket confirmation was generated on May 11th, with no human agent contact established by May 14th — eleven days into the incident. For a Max plan subscriber, who represents one of Anthropic's higher-tier paying customers, an eleven-day response window with no human escalation on an access-blocking issue points to either systemic understaffing in support operations, miscalibrated triage logic that fails to surface account lockout cases, or both. The user's consideration of a chargeback after another week of silence is a commercially rational response to being billed for a service rendered inaccessible, and it introduces potential downstream consequences for Anthropic in terms of payment processor dispute records.
The incident is instructive as a case study in the risks of federated identity systems when coverage is inconsistent across platforms. Anthropic's decision to offer Sign in with Apple as an authentication route — almost certainly required by Apple's App Store guidelines for apps that support social login on iOS — created a user population whose accounts are bound to an identity provider that Anthropic's own macOS surface does not support. Without a clear account-linking or migration mechanism, these users are structurally trapped. The broader trend in AI product development has been rapid platform expansion — native desktop apps, browser extensions, mobile apps — often outpacing the identity and account infrastructure required to support seamless cross-platform continuity. Claude's macOS authentication gap is a concrete manifestation of that pattern, where frontend deployment speed exceeds backend authentication completeness.
The episode also reflects a growing tension in AI assistant support models, where companies have aggressively deployed conversational AI for first-line customer support — in this case, the Fin bot — without adequate escalation pathways for issues that exceed bot resolution capability. Authentication and account integrity failures by definition cannot be resolved through FAQ-style bot interactions; they require backend investigation by engineers or account specialists with database access. When triage systems fail to route these cases appropriately, paying customers are left in extended limbo that erodes trust in the product independently of the underlying technical failure. Anthropic's reputation in this case is doubly exposed: once for the platform gap that caused the lockout, and again for the support infrastructure that failed to address it within any reasonable commercial timeframe.
Read original article →