Vercel adds device-bound sessions and opens AI security review to public beta
Vercel is rolling out Device Bound Session Credentials (DBSC) to help protect user sessions from replay after browser-cookie theft, while also making its AI-assisted Vercel Security Review available in public beta to Pro and Enterprise teams. The two features address different security concerns: DBSC is intended to make stolen session cookies less useful, and Security Review looks for flaws in code changes submitted through GitHub.
Binding session refreshes to a device
Infostealer malware can take browser cookies as well as passwords and other authentication credentials. A session cookie is particularly attractive to an attacker because it can represent a user’s already-authenticated session, including one established after multi-factor authentication. Replaying a stolen cookie can therefore allow access without entering the password or completing the second factor again.
DBSC is designed to limit that pass-the-cookie risk by linking a session to the device where it was created. The browser generates a public and private key pair, with the private key held in secure device hardware, such as a Trusted Platform Module (TPM) or Secure Enclave. The server issues short-lived cookies that the browser must periodically refresh by demonstrating possession of that private key.
In this model, possession of a copied cookie alone is not enough to keep the session active: the attacker would also need the device-held private key to refresh it. The approach does not prevent malware from stealing cookies, but it is intended to reduce the value and usable lifetime of a stolen cookie when the attacker lacks the associated key.
Vercel says it is introducing DBSC gradually. The technology is new and not yet broadly available, and the company had to address special cases, including OAuth applications. Google says that when a device does not support secure key storage, DBSC falls back to standard behavior rather than disrupting authentication.
Current support is limited to Chrome on Mac and Windows, according to Vercel CTO Malte Ubl. Vercel is working on a polyfill for other browsers, but Ubl cautioned that this alternative is less secure and said broader browser adoption would be desirable. The company is also working on Demonstrating Proof-of-Possession (DPoP) for its command-line interface. The announcement did not provide a release date for that work.
Security review checks changes on GitHub
Separately, Vercel Security Review is now in public beta for Pro and Enterprise teams. It is part of Vercel Agent and checks code when users push changes to GitHub. The review examines what changed since the branch’s previous deployment, considers surrounding code, traces how the changed code can be reached, and checks for existing protections before reporting a finding.
Results are delivered as comments on pull requests, through Vercel’s Security dashboard, and via the command-line interface. Vercel says each finding includes an explanation of the issue and remediation guidance, with links to the relevant commit and the deployment where it first appeared. Users can ask Vercel Agent to verify a finding and commit a minimal fix, including updated tests, or provide the finding details to another coding agent.
Reviews are triggered by GitHub push events. Vercel said use of Security Review would be free through November 10, 2026. The announcement does not specify how the service will be priced after that date.
Operational considerations
For security teams, DBSC’s browser and hardware requirements matter when assessing how session protection will work across an organization’s endpoint fleet. The fallback behavior means unsupported devices continue using standard authentication behavior, so teams should not assume that every session receives device binding. Browser compatibility and the status of Vercel’s gradual rollout are also relevant to deployment planning.
Security Review adds an automated review point to GitHub-based development workflows, but its findings still need to be handled as part of an established code-review and remediation process. Teams can evaluate how its pull-request comments, dashboard entries, and CLI output fit with existing security checks, and decide how to verify suggested fixes before merging them.
