The driving license information of thousands of Uber drivers was stolen, but it wasn’t Uber itself that was targeted in this attack.
The incident involved Fireworks AI, a company that helps other businesses implement AI models. Hackers gained access to Fireworks, obtained its internal credentials, and used these credentials to retrieve the Uber data stored there.
In the notification issued by Uber, it was reported that the driver’s license information of 3,430 people in South Carolina, 2,567 people in Massachusetts, and 1,873 people in Rhode Island had been stolen. Drivers in Illinois and California also received similar notifications, although the exact number of affected individuals has not yet been disclosed.
Uber received the notification on October 9th and immediately severed its data connection with Fireworks. The company stated that, so far, no misuse of the data had been detected and offered free credit monitoring services to those who were affected.
Fireworks pointed the finger at a criminal gang, claiming that their goal was to resell access rights to AI models. During the same attack, data from Notion was also stolen; Cursor also mentioned this week that its own data breach initially stemmed from a key that was exposed by Fireworks.
The lessons learned from this incident are more concrete than simply the issue of data leakage: when a company entrusts its data to an inference service provider for model development, it is essentially extending its own security boundaries into the provider’s system. If one supplier encounters problems, several downstream companies will be affected as well. Companies that use third-party inference platforms would do well to check this week whether the permissions granted to these providers are too extensive.