Uber Drivers’ Licence Data Stolen Through AI Vendor Fireworks, the Second Big Breach Through It This Month

Uber is writing to thousands of its US drivers to tell them that their personal information, including driver’s licence details, was stolen in a breach at one of its technology providers. TechCrunch has identified the provider as Fireworks AI, a company that runs open AI models for developers, and Uber confirmed it to the publication. It is the second big-name customer, after Notion, known to have lost data through Fireworks this month. Uber’s notice says its own systems were not the route in: the attacker used stolen credentials belonging to the provider to get into Uber’s account with it.

#What Uber’s letter says

Uber’s notification letter, dated Tuesday 10 November and filed with California’s attorney general, sets out the timeline:

• 9 October: a service provider told Uber that an unauthorised third party had compromised the provider’s internal credentials and used them to access Uber’s account with it.
• Uber says it “immediately severed data connections with the service provider” and began investigating.
• 13 October: Uber’s investigation found the attacker had used those credentials to access Uber data connected to the provider.
• Uber is offering affected people credit monitoring and fraud assistance through Cyberscout, a TransUnion company.

The template filed in California leaves the data fields blank. TechCrunch, which reviewed state filings, reports driver’s licence information for 2,567 people in Massachusetts, 3,430 in South Carolina and 1,873 in Rhode Island, with numbers in Illinois and California not given. Uber has not said how many people are affected in total, and TechCrunch expects the figure to rise. The notices so far are to US drivers; we have seen nothing saying riders, or drivers outside the US, are affected.

#The vendor behind two breaches this month

BleepingComputer reports that Fireworks AI confirmed an attacker reached its development environment through a third-party cloud provider on 23 October, and that it had a separate breach in early October in which attackers accessed environment variables holding API keys and tokens. Notion disclosed on 11 November that Fireworks’ breach exposed Notion AI chat transcripts for a small number of workspaces, even though, according to Notion, Fireworks had committed to keep none of its data. TechCrunch reports that Fireworks attributed its breach to a financially motivated group that sells stolen access tokens for AI models.

Fireworks confirmed to TechCrunch that it had found unauthorised access to some systems affecting customer data, but did not answer its questions.

#Our view

We think this is the clearest example yet of a risk that most organisations have not mapped: the companies that run AI models for your products are now holding your data, and they are being breached like any other supplier. Uber and Notion did not lose data through a flaw in an AI model. They lost it because an AI vendor kept their data and had its own keys stolen, which is an ordinary supply-chain failure in a new place.

It is the same pattern as the Expedia breach we covered this week, where trusted access to a data platform was the way in. In both cases the question for IT and security teams is not “is this AI safe?” but “what have we given this supplier, and do they still need it?”

We would also question the “zero data retention” promises that AI vendors make. Notion says it had one from Fireworks; its users’ transcripts were still in the breached data. Treat a retention promise as a contract term to audit, not a technical guarantee.

#What to do

1. List your AI vendors, including the ones behind features in products you buy, and find out what data each one holds and for how long.
2. Ask for evidence, not policy, on data retention: when data is deleted, and how they would know if it were not.
3. Rotate keys you have given AI services if you have any doubt about the vendor, and scope them to the minimum.
4. If you drive for Uber in the US and receive a letter, enrol in the credit monitoring within the 90 days the letter allows, and be wary of messages that use your details to look official.