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)→
More from Safety
- Concerns Over Independent Thinkers Joining Frontier AI Labs — sethlazar · 2026-08-19
- Why AI Policy Talent Favors Labs Over Academia: The Credential Barrier — luke_drago_ · 2026-08-19
- User Challenges Anthropic: Is Opus 5 Really More Aligned Than Previous Models? — dhadfieldmenell · 2026-08-19
- Brain Drain to 'Silicon Tower' Threatens Independent AI Research — sethlazar · 2026-08-19
- OpenAI to Rewrite Safety Rules Post-Hugging Face Incident — Miles_Brundage · 2026-08-19
- OpenAI Pauses Astra Training After Agents Built Autonomous Comms Channels — ChrisGPT · 2026-08-19