Anthropic describes sandboxing for Claude Code that restricts file writes to permitted directories and routes network traffic through a proxy that only allows approved hosts, so a prompt-injected agent cannot modify sensitive files or exfiltrate data freely. Anthropic says internal use showed an 84% reduction in permission prompts, and it released the sandbox runtime, built on bubblewrap and macOS seatbelt, as an open-source research preview.
Why it matters
It is a concrete containment control that limits the blast radius of prompt injection in coding agents regardless of model behavior.
Key facts
As stated in the sources, with where to find them.
- Anthropic reports sandboxing safely reduced permission prompts by 84% in internal usage.Main post
- Runtime released as anthropic-experimental/sandbox-runtime, using Linux bubblewrap and macOS seatbelt.Open source section
Findings that cite this record
Key questions this bears on
- Do cyber evaluations of AI agents stay contained?Not reliably. Agents have reached real systems from evaluation, training and research; red-teaming found containment gaps OpenAI says it largely fixed.
- Can AI agents defend and oversee systems on their own?Not reliably. Some designs improve benchmark triage, but autonomous defense lacks operational validation and production oversight of agents has documented gaps.
Sources
Related records
Nov 20, 2025
Jul 28, 2025
Oct 29, 2026
Sep 22, 2026
Aug 4, 2026
Jun 3, 2026