Console shows repeated AI/LLM endpoint activity (AI/LLM ENDPOINT DETECTED) and numerous external requests to softgen.ai. Automatic calls to LLM endpoints on load can expose user data or app state to a third party without explicit user consent.

If such calls are essential and properly consented/handled, this may be legitimate; however the logs imply lack of explicit consent gating and visibility.

Gate LLM/AI calls behind explicit user consent, provide a clear privacy policy, and consider moving calls behind a server-side proxy or implement data minimization. Add retry/backoff and proper error handling instead of unconditional calls.

Enhances user privacy, aligns with data protection expectations, and reduces risk of unintended data exposure.

The image URL exposes an internal user identifier (e.g., /users/4.jpg) directly in the request path and in the _next/image URL parameter. This can enable user identification or correlation across requests and layers, constituting unnecessary exposure of user identifiers in URLs.

Some apps use numeric user IDs in internal paths; if those IDs are not publicly linked to real-world identities and are not exposed beyond authenticated contexts, risk is lower. However, URLs are often cached and logged, increasing exposure risk.

Replace direct user IDs in URLs with opaque tokens or asset IDs that do not reveal the user. Serve user-specific assets via server-side mediation that requires authentication and does not leak IDs in the URL. If exposing IDs is required, redact or map to non-identifying tokens before sending in URLs.

Prevents leakage of potentially identifying user IDs in client-facing URLs, reducing risk of correlation and profiling across sessions and services.

A POST request is sent to Sentry's ingest endpoint (third-party error tracking). Error payloads can include stack traces, environment data, and potentially user identifiers or IP addresses. Without explicit privacy safeguards or user opt-in, this risks exposing personal data to a third party.

Sentry usage is common for error monitoring; if the implementation redacts PII, obfuscates IPs, and user data is not included, this would be acceptable. The provided data does not reveal the payload, so assume privacy controls may be absent or insufficient.

Enable PII redaction in Sentry payloads, anonymize IP addresses, and disable default collection of user-identifying data. Add a clear privacy notice and/or user consent for error telemetry. Consider using a privacy-friendly alternative or self-hosted error collection. Implement a before_send hook to scrub sensitive fields.

Protects user privacy by preventing leakage of personal data through error reporting and aligns with data minimization principles and regulations.