Sandlock: unprivileged Linux sandbox for AI agent code
Start here
Yusheng Zheng and co-authors published a paper introducing Sandlock, a lightweight Linux process sandbox for confining untrusted AI agent code. Its design splits policy in two: static, input-independent rules are compiled into kernel-enforced rules, while a narrow supervisor handles runtime-dependent decisions and virtualized effects. The authors say this lets Sandlock enforce filesystem, network, IPC and syscall policies without root, cgroups, images or mandatory namespaces. On the authors' workstation, Sandlock adds roughly 5 ms of startup overhead and runs Redis at bare-metal throughput, within measurement noise. A pipeline operator also supports per-stage confinement, separating data, network and untrusted-content capabilities. These performance figures are the authors' own measurements.
Generated from reports · updated 20 h ago
Report timeline
Follow the reports to see the event from different sides.
- Hacker News · AISandlock: Confining AI Agent Code with Unprivileged Linux Primitives
Sandlock is a lightweight Linux process sandbox that confines untrusted AI agent code without root, cgroups, images, or mandatory namespaces. It compiles static, input-independent policy into kernel-enforced rules while a narrow supervisor handles runtime-dependent decisions, enforcing filesystem, network, IPC, and syscall policies. It adds roughly 5 ms of startup overhead and runs Redis at bare-metal throughput.
Heat for this event
- Current, comparable range
- 6
- Peak, comparable range
- 10November 15 21:00
- Change in 24 hours
- –
The trend compares the same participants over hours observed continuously, so its range can be narrower than the current heat count. Hover or tap the chart for hourly heat; the arrow keys move through it.