Notion User Data Exposed Following Breach at AI Provider
A security failure at a third-party partner has led to the exposure of sensitive user chat logs.
Notion, a central tool for managing workflows for millions of users globally, has reported a significant security incident. The company announced that a cyberattack on one of its primary AI service providers, Fireworks AI, led to the leakage of Notion user chat transcripts.
What happened at Fireworks AI?
The incident began on 23 October, when attackers gained entry into the Fireworks AI development environment through a third-party cloud service. Hackers accessed resource metadata, billing information, and model weights. Notably, this marks the second security breach for Fireworks AI recently; in early October, the company reported a leak of API keys and tokens.
Notion became aware of the incident on 2 November. Initially, the information suggested that only synthetic, test data was involved. However, after it was confirmed on 8 November that actual user data had been compromised, Notion immediately severed its connection with Fireworks AI, deleted all data from the partner's systems, and halted model training processes.
What data was exposed?
According to Notion, the leak affected the 'Insights' feature, which allows managers to analyse how their teams use artificial intelligence. The company stated that the number of affected users represents less than 1% of Notion’s total user base.
Importantly, the majority of the leaked chats were 'unidentifiable,' meaning they did not contain information about the author or a specific workspace. Nevertheless, Notion emphasised that Fireworks AI was contractually prohibited from storing this data, which makes the incident particularly critical.
How to protect yourself?
If your workspace was affected, Notion will send you a notification. For users who receive an alert, the company advises checking their data through the 'Security incident review' section in the desktop web interface. It is also recommended to change any passwords or API keys that may have been mentioned in the chats.
If you or your team administrator have not received a notification, your data is likely secure and no further action is required. This case serves as a reminder that when using digital tools, it is always important to verify data privacy policies and the reliability of the third-party services to which our data is transmitted.