Vercel launches $1M open challenge to test AI sandbox security guardrails

ccerrato147 · x · 2026-08-19

Vercel is putting $1M towards open security verification for Vercel Sandbox, inviting anyone to test any model to attempt an escape. The challenge targets the Firecracker microVM and host-side network boundaries, offering up to $50k per report via HackerOne. The goal is to bring transparency to real-world guardrail exploitability of frontier models and share findings to strengthen global cybersecurity.

Related event: Vercel Offers $1M Bounty to Break Its AI Agent Sandbox(2 posts)→

Original post →

More from Safety

Safety channel →