Glossary
Sandbox escape
Breaking out of the isolation that is supposed to keep one tenant's code, or one agent's actions, from reaching the host or other tenants.
Railway's October breach began with a partial filesystem-isolation escape on a shared build VM that exposed a build credential. The same class of flaw matters for AI agents that execute code: an agent sandbox is only as strong as the boundary and the credentials visible inside it. Per-task ephemeral VMs with narrowly scoped credentials are the standard fix.
Related terms
Get the brief in your inbox
One email every morning: the stories that mattered in AI, each with a one-line "why it matters" for your role. Free, no ads, unsubscribe any time.
Published daily by AI Briefing. We never share your address.