Vercel has disclosed new findings in its ongoing investigation into a recent security incident, confirming that more customer accounts were affected than initially believed.
Expanded Investigation Reveals More Impact
The company reported that after broadening its analysis, including reviewing network activity and environment variable access logs, it discovered an additional group of accounts showing signs of unauthorized access.
Alongside this, Vercel also identified a separate set of customer accounts that had been compromised prior to the incident. These earlier breaches may have resulted from tactics such as social engineering, malware infections, or other attack methods unrelated to the main breach.
Affected users have been notified, although the company has not shared the exact number of impacted accounts.
Breach Linked to Context.ai Compromise
The root cause of the incident traces back to a compromise involving Context.ai. Attackers gained access through an account used by a Vercel employee, which allowed them to take control of the employee’s Google Workspace account and subsequently access internal Vercel systems.
From there, the attacker moved laterally within the environment, identifying and decrypting certain non-sensitive environment variables.
Malware Infection Identified as Possible Entry Point
Further analysis by Hudson Rock suggests that the breach may have originated from a malware infection involving Lumma Stealer.
The infection reportedly occurred when an employee associated with Context.ai downloaded malicious files while searching online for unauthorized gaming tools. This initial compromise is believed to have triggered a chain of events leading to the broader attack.
Broader Threat Activity Observed
According to Vercel leadership, the attacker’s activity was not limited to a single organization. Threat intelligence indicates that malware was distributed more widely, targeting systems in search of valuable authentication tokens and credentials tied to cloud platforms and developer services.
Concerns Around Unauthorized AI Tool Usage
The incident has also raised questions about the use of third-party AI tools within organizations. It remains unclear whether the use of Context.ai’s office suite by the employee was officially approved or an example of “shadow AI,” where tools are used without proper security review.
Such practices can introduce hidden risks, especially when integrated systems rely on OAuth permissions that inherit user-level trust.
Security Implications of OAuth Abuse
Experts warn that while OAuth integrations simplify access, they can also be exploited if compromised. Attackers leveraging trusted integrations may bypass traditional security controls, making detection more difficult.
Security analysts emphasized that the speed and efficiency of the attacker’s actions were particularly notable. Instead of focusing solely on data theft, the attacker rapidly mapped internal environments, increasing the potential impact before detection.
Found this article interesting? Follow us on X (Twitter) , Facebook, Blue sky and LinkedIn to read more exclusive content we post.
