Vercel Breach: Detecting Account Compromise via Context.ai Internal System Intrusion

On Wednesday, Vercel disclosed a concerning expansion of a security incident involving unauthorized access to its internal systems. This breach, linked to third-party integrations associated with Context.ai, has resulted in the compromise of additional customer accounts beyond the initially identified scope.

For defenders, the urgency here is twofold: the attacker bypassed standard perimeter controls by leveraging internal system access, and the investigation is still evolving. This isn't just a phishing campaign; it is a supply-chain and credential-theft scenario where the threat actor potentially possesses the ability to manipulate customer environments. If your organization relies on Vercel for hosting or CI/CD, you must assume that access tokens, environment variables, and deployment secrets may have been exposed.

Technical Analysis

Affected Platform: Vercel (Cloud Platform for Frontend/Fullstack)

Attack Vector: The incident leveraged unauthorized access to Vercel's internal systems, likely through compromised credentials or API tokens linked to a third-party integration (Context.ai). This allowed the attacker to pivot from internal infrastructure to customer tenant data.

Attack Chain (Defender Perspective):

1. Initial Access: Attainer gains access to Vercel internal tools/network (Context.ai linked vector).
2. Lateral Movement: Internal access used to query or impersonate customer accounts/API keys.
3. Objective: Unauthorized access to customer code repositories, environment variables, and deployment histories.

Exploitation Status: Confirmed Active Exploitation. Vercel has confirmed additional accounts were identified by reviewing network requests and environment indicators.

Detection & Response

This incident requires hunting for anomalies in API usage, environment variable modifications, and unexpected deployment activities. Since the attacker had internal access, they may have accessed data that doesn't trigger standard "failed login" alerts.

SIGMA Rules

Detect suspicious modifications to environment variables and unexpected project access patterns within Vercel logs (ingested via SIEM).

---
title: Vercel Environment Variable Modification
id: 8a2b4c9d-1e3f-4a5b-8c6d-7e8f9a0b1c2d
status: experimental
description: Detects the creation or modification of environment variables in Vercel, which may indicate an attacker injecting exfiltration endpoints or stealing secrets.
references:
  - https://vercel.com/docs/security
author: Security Arsenal
date: 2026/04/16
tags:
  - attack.credential_access
  - attack.t1552.001
logsource:
  product: vercel
  service: audit
detection:
  selection:
    action|contains:
      - 'env.create'
      - 'env.update'
      - 'env.remove'
  filter_legit:
    actor|startswith:
      - 'known_service_account'
  condition: selection and not filter_legit
falsepositives:
  - Legitimate developers updating configuration
level: high
---
title: Vercel Impersonation or High Privilege Access
id: 9b3c5d0e-2f4g-5b6c-9d7e-0f1a2b3c4d5e
status: experimental
description: Detects access to Vercel internal administrative interfaces or impersonation actions often associated with supply chain breaches.
references:
  - https://vercel.com/docs/security
author: Security Arsenal
date: 2026/04/16
tags:
  - attack.initial_access
  - attack.t1078
logsource:
  product: vercel
  service: audit
detection:
  selection:
    action|contains:
      - 'impersonation.start'
      - 'team.audit'
      - 'internal.access'
  timeframe: 24h
  condition: selection
falsepositives:
  - Authorized internal audits (rare)
level: critical

KQL (Microsoft Sentinel)

Hunt for anomalous deployment or environment changes in the last 7 days. This assumes Vercel logs are routed to a custom table (e.g., VercelAuditLogs) or Syslog.

// Hunt for unexpected environment variable changes
let KnownIPs = _GetWatchlist('Vercel_Known_IPs');
VercelAuditLogs
| where TimeGenerated > ago(7d)
| where action in~ ("env.create", "env.update", "deployment.create")
| extend SourceIp = tostring(coalesce(ip_address, sourceIP))
| where SourceIp !in (KnownIPs) or isempty(SourceIp)
| project TimeGenerated, action, userId, projectName, SourceIp, userAgent
| order by TimeGenerated desc

Velociraptor VQL

Hunt developer workstations for signs that the Vercel CLI was used to dump environment variables or list teams, which might indicate post-exploitation recon if the developer's local token was compromised.

-- Hunt for Vercel CLI configuration access or secret dumps
SELECT FullPath, Mtime, Size, Mode
FROM glob(globs="/*/.vercel/config.")
WHERE Mtime > now() - 7d
-- Alternative: Hunt for process execution of vercel cli with env listing
SELECT Pid, Name, CommandLine, Exe
FROM pslist()
WHERE Name =~ "node.exe"
  AND CommandLine =~ "vercel" 
  AND CommandLine =~ "env" 
  AND Exe NOT =~ "Program Files"

Remediation Script (PowerShell)

Run this on developer machines to ensure the local Vercel session is revoked and identify recent suspicious activity.

# Function to check Vercel CLI status and list recent projects
function Check-VercelSecurity {
    Write-Host "Checking Vercel CLI status..."
    
    # Check if Vercel CLI is installed
    $vercelPath = Get-Command vercel -ErrorAction SilentlyContinue
    
    if ($vercelPath) {
        Write-Host "[+] Vercel CLI found at: $($vercelPath.Source)"
        
        # Identify current logged in user
        $whoami = vercel whoami 2>&1
        if ($LASTEXITCODE -eq 0) {
            Write-Host "[!] Currently logged in as: $whoami"
            Write-Host "[ACTION REQUIRED] Please run 'vercel logout' to force re-authentication."
        } else {
            Write-Host "[+] No active Vercel session found."
        }

        # List local config files for review
        $configPath = "$env:USERPROFILE\.vercel"
        if (Test-Path $configPath) {
            Write-Host "[+] Found Vercel config directory: $configPath"
            Write-Host "[Review] Please manually inspect $configPath\config. for unknown tokens."
        }
    } else {
        Write-Host "[-] Vercel CLI not found on this machine."
    }
}

Check-VercelSecurity

Remediation

1. Force Token Rotation: Assume all Vercel API tokens and deployment secrets created before the incident date (April 2026) are compromised. Revoke these immediately in the Vercel Dashboard Settings > Tokens.
2. Audit Integrations: Review all OAuth applications and Git integrations connected to your Vercel account. Specifically, investigate and temporarily revoke access related to Context.ai or any third-party tools you do not explicitly recognize.
3. Review Environment Variables: Check the "Environment Variables" tab for all projects. Look for unknown variables that might be used for data exfiltration (e.g., unexpected webhook URLs or foreign proxy endpoints).
4. Inspect Deployment Logs: Analyze git commit history and deployment logs for any code or deployments pushed from unverified IP addresses or unusual timestamps during the breach window.
5. Enable MFA: Ensure Multi-Factor Authentication is enforced for all team members on the Vercel account.

Official Vendor Advisory:

• Vercel Security Update

Related Resources

Security Arsenal Incident Response Services AlertMonitor Platform Book a SOC Assessment incident-response Intel Hub