Uber notifies thousands of drivers after breach at AI vendor Fireworks AI
The ride-hailing company is sending state-mandated breach notices after hackers stole driver's-license data through a shared AI infrastructure provider that also exposed chat data at Notion.

Uber Technologies has begun telling thousands of its drivers that their personal information, including driver's license numbers, was exposed after hackers broke into systems run by Fireworks AI, a San Francisco artificial-intelligence infrastructure company that Uber uses to operate machine-learning models. The notifications, filed with state regulators over the past two weeks and first detailed in a TechCrunch report published Friday, make this the second data-security incident Uber has disclosed in 2026 and place it among several companies swept up in a breach at a shared AI vendor.
Uber told TechCrunch the incident traces back to "a security incident impacting one of our third-party vendors" and referred detailed questions to Fireworks AI, which confirmed unauthorized access to some of its systems but declined to answer specific questions about the intrusion. According to the company, Fireworks notified Uber on October 9 that attackers had stolen internal credentials from its systems and then used those credentials to reach data Uber had stored with the vendor.
The numbers so far
The scale of the breach is still coming into focus through a patchwork of state filings, a reporting requirement that varies from state to state and tends to surface slowly. A sample breach notification Uber filed with the California Attorney General's office lists the date of the breach as October 7, 2026, with Uber's report reaching the state on November 10. Massachusetts regulators recorded 2,567 affected residents; South Carolina logged 3,430; and Rhode Island counted 1,873, according to filings reviewed by TechCrunch. Illinois has an entry pending a final count, and California drivers were notified without a published total. Uber has not released an overall figure, and because notification deadlines differ by state, the running total is expected to climb as more jurisdictions post their own filings in the coming weeks.
A vendor breach that became everyone's problem
Fireworks AI sells access to open-weight large language models and the infrastructure to run them, a niche that has grown quickly as companies look for cheaper alternatives to building their own AI systems. That role made it a single point of failure once attackers got in. Fireworks disclosed two separate intrusions this fall: an early-October incident in which hackers accessed environment variables containing API keys and access tokens for a limited number of accounts, and a second breach on October 23 in which an attacker reached Fireworks' development environment through a third-party cloud provider, exposing resource metadata, billing information, model weights and customer datasets.
Fireworks has attributed the intrusions to a financially motivated group that resells stolen access tokens for large language models, rather than to a state-linked actor, though no group has publicly claimed responsibility. The episode illustrates a risk that security agencies have been warning about as companies race to bolt AI features onto their products: a single compromised vendor can cascade into breaches at every customer that fed it data. The Cybersecurity and Infrastructure Security Agency has repeatedly flagged this kind of software supply-chain exposure, noting that vulnerabilities can enter at any stage of a vendor relationship and that "a supply chain is only as strong as its weakest link."
Notion and others caught in the same breach
Uber is not the only Fireworks customer affected. Notion, the workplace productivity company, disclosed in a security advisory reported by BleepingComputer on November 11 that the same breach exposed AI chat transcripts tied to its "Insights" feature, which lets workspace administrators see how employees use Notion's AI tools. Fireworks runs some of the open-weight models behind that feature and had been storing related conversation data. Notion said it was alerted to an exposed dataset on November 2, cut off its traffic to Fireworks and deleted its data from the vendor's systems within a day, and received confirmation on November 8 that customer data, including its own, had been accessed.
Notion said fewer than 1% of its users were affected, that most of the exposed transcripts could not be tied to a specific person or workspace, and that none contained the results of AI tool calls. The company is notifying affected workspace owners individually and advising anyone who entered passwords or API keys into an affected chat to rotate those credentials. Uber's driver notifications, by contrast, involve more sensitive identity documents rather than chat logs, which is why state privacy laws require the formal notices now surfacing in regulators' databases.
Uber's drivers are the clearest group harmed here: thousands of gig workers whose state-issued identification numbers are now in the hands of unknown attackers, with the standard exposure to identity theft and fraudulent-account risk that follows. The Federal Trade Commission's guidance for breached businesses recommends offering affected individuals at least a year of credit monitoring when sensitive identification data is exposed, though Uber's filings reviewed so far do not specify what remedy, if any, it is offering drivers. Notion's exposure is narrower and framed by the company as low-risk, since the transcripts it lost were largely anonymized by the time they reached Fireworks' systems.
What happens next
More state notifications are likely in the coming weeks, since large multistate breaches typically trickle out as each attorney general's office posts its own filing on its own timetable; Illinois alone has yet to post a final figure. Uber has not said whether it plans to drop Fireworks as a vendor, and Fireworks has not published a detailed post-incident report beyond its acknowledgment that customer data, including model weights and datasets belonging to some clients, was accessed. This is Uber's second disclosed breach of the year, following an earlier incident involving its freight unit in which an extortion group claimed responsibility for stealing data, a pattern that is likely to draw fresh scrutiny of how thoroughly gig-economy and technology companies vet the AI vendors now embedded deep in their internal systems.

OpenAI Lets Outside Researchers Embed Inside Its Labs to Hunt for Rogue AI Behavior
Meta, Signal and Tailscale tell Canadian senators Bill C-22 would force them to build surveillance tools

Alibaba Launches Qwen4-Max, Betting Lower Prices Can Challenge US AI Leaders
