Detailed Analysis
A Claude user has reported a persistent and platform-agnostic failure in which the AI assistant stops accepting new messages entirely, across both existing conversations and newly initiated ones. The symptom manifests consistently: typed messages appear in the chat interface but elicit no response from Claude, the Retry button produces no effect, and refreshing the page causes the failed message to disappear without resolution. Critically, the user has replicated the issue across multiple devices and operating systems — Chrome on PC, MacBook, iPhone, and via dedicated iOS and Windows desktop apps — ruling out device-specific or browser-specific causes.
The troubleshooting steps undertaken by the user were extensive and methodical. Cache clearing, full logout across all devices, and even revoking Claude's OAuth connection to Google were all attempted without success. The involvement of a competing AI assistant (Gemini) in diagnosing the issue is a notable detail: Gemini suggested the root cause may be an "Account DB Sync Failure," pointing to a backend data integrity problem on Anthropic's infrastructure rather than anything client-side. This framing, while coming from a third-party AI system and carrying no official weight, is nonetheless consistent with the behavioral pattern described — a user account that is authenticated and functional at the surface level but silently broken at the data persistence layer.
The user's observation that the issue began shortly after requesting and downloading a copy of their personal data is a significant potential clue. Many platforms implement data export pipelines that interact directly with user account records and database entries. If Anthropic's data export process contains a bug that introduces inconsistencies into user account state — such as corrupted session tokens, malformed conversation metadata, or broken database references — it could plausibly render an account unable to persist new messages, which aligns precisely with the described symptoms. The failure mode, where messages appear to submit but are silently dropped, is characteristic of a backend write failure rather than a frontend or authentication error.
This incident highlights a broader reliability concern for AI assistant platforms as they scale: the coupling between ancillary account management features (like GDPR-compliant data export tools) and core conversational functionality. When these systems share underlying database infrastructure, a bug in one pipeline can cascade into complete service unavailability for affected users. The fact that no standard self-service remediation step resolved the issue — and that the user was ultimately directed to contact support directly — suggests the problem requires manual backend intervention, raising questions about Anthropic's tooling for diagnosing and recovering individual account states at scale. As Claude's user base grows, the robustness of such account-level infrastructure will become increasingly critical to user trust and retention.
Read original article →