99Ai
Monthly
Unauthenticated attackers can export all users' chat logs from 99AI 4.3.0 and earlier by sending a POST request to the /api/chatLog/exportExcel endpoint, which lacks any authentication guard (CWE-306). The vulnerability allows remote retrieval of xlsx files containing usernames, emails, prompts, and AI answers without credentials or user interaction, and affects any network-reachable instance with no special configuration. No public exploit code has been identified at time of analysis, but the exposure is high-risk because it leaks sensitive conversation data.
99AI through version 4.3.0 allows unauthenticated attackers to read other users' private chat messages and task data, because the querySingleChat handler in the chatLog service performs no ownership check on the requested chat identifier. A remote attacker simply supplies a fingerprint header to be issued a low-privilege visitor JWT, then enumerates sequential integer chatId values to harvest message content belonging to other users. The source CVSS 4.0 vector rates this 6.9 (Medium) with PR:N/UI:N and low confidentiality impact, and our independent assessment agrees it is trivially and reliably exploitable, though it is read-only, grants no write or administrative access, and is not confirmed actively exploited (no CISA KEV listing) or accompanied by packaged exploit code; a public technical writeup describing the attack path is referenced.
Unauthenticated attackers can export all users' chat logs from 99AI 4.3.0 and earlier by sending a POST request to the /api/chatLog/exportExcel endpoint, which lacks any authentication guard (CWE-306). The vulnerability allows remote retrieval of xlsx files containing usernames, emails, prompts, and AI answers without credentials or user interaction, and affects any network-reachable instance with no special configuration. No public exploit code has been identified at time of analysis, but the exposure is high-risk because it leaks sensitive conversation data.
99AI through version 4.3.0 allows unauthenticated attackers to read other users' private chat messages and task data, because the querySingleChat handler in the chatLog service performs no ownership check on the requested chat identifier. A remote attacker simply supplies a fingerprint header to be issued a low-privilege visitor JWT, then enumerates sequential integer chatId values to harvest message content belonging to other users. The source CVSS 4.0 vector rates this 6.9 (Medium) with PR:N/UI:N and low confidentiality impact, and our independent assessment agrees it is trivially and reliably exploitable, though it is read-only, grants no write or administrative access, and is not confirmed actively exploited (no CISA KEV listing) or accompanied by packaged exploit code; a public technical writeup describing the attack path is referenced.