Uber drivers’ licence data stolen in Fireworks AI hack

Uber has confirmed a significant data security breach involving a third-party vendor, Fireworks AI, which resulted in the unauthorized exfiltration of sensitive personal information belonging to thousands of the ride-sharing platform’s drivers. The incident highlights the growing risks associated with the interconnected supply chain of artificial intelligence service providers, where a single vulnerability in a vendor’s infrastructure can ripple across multiple high-profile clients.
According to reports, the breach originated when hackers gained access to internal credentials belonging to Fireworks AI. These credentials served as a gateway, allowing malicious actors to bypass security protocols and reach data stored by the AI infrastructure firm on behalf of its partners, including Uber. While the full scope of the breach remains under investigation, filings with state attorneys general across the United States have begun to paint a sobering picture of the scale of the exposure.
Chronology of the Breach
The timeline of the event, as reconstructed through regulatory filings and public disclosures, began in early October 2026. Fireworks AI, a high-growth startup valued at $17.5 billion as of July 2026, disclosed that it experienced two distinct security incidents during that month.
On October 9, 2026, Fireworks AI notified Uber that an unauthorized party had successfully compromised internal credentials. This initial intrusion served as the foundation for the subsequent data theft. Investigations revealed that the attackers utilized these elevated privileges to pivot into Uber’s datasets, which had been integrated into the Fireworks AI ecosystem for model training or inference purposes.
By mid-November, the impact began to manifest in official legal notifications. Uber initiated mandatory reporting to state regulators, confirming that driver’s licence information and other personal identifying details were compromised. Shortly after, it was revealed that Fireworks AI had suffered a second, separate breach, this time involving a compromise at a third-party cloud service provider that allowed attackers to access its development environment.
The Scale of the Exposure
While Uber has yet to release a definitive, global count of affected drivers, the fragmented disclosures provided to state officials provide a preliminary snapshot of the damage. As of the latest reporting cycle, the following figures have been confirmed:
- Massachusetts: 2,567 affected individuals
- South Carolina: 3,430 affected individuals
- Rhode Island: 1,873 affected individuals
When combined with ongoing notifications in California and Illinois—where specific victim counts have not yet been disclosed—the total number of impacted drivers is expected to climb well into the five-figure range. The inclusion of driver’s licence data is particularly concerning for security experts, as this information is a primary component used by identity thieves to open fraudulent financial accounts or bypass "know your customer" (KYC) verification protocols on other platforms.
The Rise of LLMjacking and Third-Party Risk
The incident has been linked by security researchers to a phenomenon known as "LLMjacking." This practice involves attackers targeting companies that provide access to large language models (LLMs) or AI inference services. By stealing API keys or internal administrative credentials, hackers can effectively "hijack" the computing power and data access of these platforms for their own purposes, such as training their own models or harvesting proprietary datasets.
In the case of Fireworks AI, the attackers did not stop at billing data. They reportedly accessed sensitive "secrets"—cryptographic keys or passwords used to secure digital infrastructure—as well as datasets and unfinished model weights. Other companies utilizing Fireworks AI, such as Notion, have also confirmed that their services were impacted, with Notion noting that AI chat transcripts from approximately 1% of its workspaces were exposed during the breach.
The irony of the situation has not been lost on industry analysts. Uber, a company that prides itself on advanced algorithmic logistics and robust security architecture, found itself vulnerable through a vendor it trusted to manage its AI integration. The industry-wide push to adopt generative AI models has forced companies to share data with a burgeoning ecosystem of "inference providers"—firms that host the models and process the queries. If these providers do not maintain the same level of security posture as their enterprise clients, the entire chain of custody for sensitive data is compromised.
Official Responses and Legal Implications
Uber has maintained a measured, if brief, response to the incident. In its communications with drivers, the company has emphasized that the breach occurred on the vendor’s side and that it is working with law enforcement and security consultants to mitigate the damage. However, the company has declined to provide further public detail regarding why sensitive driver’s licence data was being processed or stored by an AI inference provider in the first place, citing the ongoing nature of the investigation.
Fireworks AI, for its part, has faced intense scrutiny regarding its security protocols. The company has not provided detailed answers to inquiries regarding the specific security gaps that allowed the initial credential theft to occur. Given its valuation and the critical nature of the services it provides to the AI development community, the incident raises questions about the maturity of security governance at high-valuation, high-growth AI startups.
Legal experts suggest that Uber could face increased regulatory scrutiny under various data privacy laws, including the California Consumer Privacy Act (CCPA) and the General Data Protection Regulation (GDPR), if applicable. Even when a breach occurs at a third party, the primary data controller—in this case, Uber—often bears the ultimate responsibility for ensuring that the third party is a suitable steward of the data.
Broader Industry Impact and Analysis
This incident represents the second significant cybersecurity event for Uber in 2026, following a breach involving Uber Freight in August. For the broader tech industry, the event serves as a warning about the "AI supply chain." As enterprises rush to integrate AI into their operations, they are increasingly handing over vast quantities of sensitive data to vendors whose core competency is model performance rather than cybersecurity.
The implications are two-fold:
- Vendor Due Diligence: Corporations will likely be forced to implement more rigorous, and potentially more expensive, security auditing processes for AI vendors. It is no longer sufficient to treat these providers as simple software-as-a-service (SaaS) entities; they must be treated as critical infrastructure partners.
- Data Minimization: There is a growing consensus that sensitive personal information, such as driver’s licences or social security numbers, should never be part of the training or inference datasets shared with third-party AI providers. The practice of "data anonymization" must become more robust, ensuring that even if a vendor is breached, the data stolen is essentially useless to the attacker.
As the investigation continues, the focus will likely shift toward whether Fireworks AI had sufficient encryption in place for the data it held. If the data was stored in a clear-text format or protected by easily guessable credentials, the path toward class-action litigation against both the vendor and the client could become significantly clearer.
For the thousands of Uber drivers currently dealing with the fallout, the reality is more personal. The theft of identity-related documents necessitates immediate action, including the freezing of credit reports and monitoring of financial accounts for signs of unauthorized activity. While Uber has promised support, the long-term risk of identity fraud remains a significant concern for those whose data was caught in the wake of this sophisticated and far-reaching security failure. The incident serves as a stark reminder that in the age of artificial intelligence, the most advanced security systems are only as strong as the most vulnerable vendor in the supply chain.







