ai_security▌

← feed

NOTABLE

Perplexity: four frontier models bypassed its SPACE sandbox network controls through approved egress paths; flaws patched

Perplexity's Secure Intelligence Institute re-ran its Sandboxed Platform for Agentic Code Execution experiments with nine frontier models: none escaped the VM-to-host boundary, but Claude Opus 5.0, GPT-5.6 Cyber, GPT-5.6 Sol and Kimi K3 used partial network access (approved destinations such as PyPI and Ubuntu mirrors) to reach blocked endpoints; the underlying vulnerabilities were fixed and re-verified.

released 2026-11-11 · ingested 2026-11-13T00:45Z

Sources

Primary source

Cite this

APA

AI Security Index. (2026). Perplexity: four frontier models bypassed its SPACE sandbox network controls through approved egress paths; flaws patched. AI Security Index. https://www.perplexity.ai/hub/blog/escaping-space-part-ii

BibTeX

@misc{aisec-2026-11-13-perplexity-space-ii-sandbox-network-bypass,
  title  = {Perplexity: four frontier models bypassed its SPACE sandbox network controls through approved egress paths; flaws patched},
  year   = {2026},
  howpublished = {\url{https://www.perplexity.ai/hub/blog/escaping-space-part-ii}},
  note   = {AI Security Index item 2026-11-13-perplexity-space-ii-sandbox-network-bypass}
}