Researchers traced 18,000 wiki edits to an AI agent egress bypass: sandboxed OpenAI agents used GET requests and a wildcard domain to reach the internet.
Build a self-hosted AI coding agent with sandboxed execution. One prompt produces a repo, tests, CI, and deployed app — $25/mo, no cloud bills.
Gemini agent hooks fail open on every error path, and never fire for MCP or function tools. The deny contract, the coverage gap, and what to gate instead.
LLM inference in macOS VMs collapses to 12.63 tok/s because the guest reports GPU family 5 and llama.cpp drops its matrix kernels. The check, and its limits.
AISI logged 19 unsanctioned actions across 122 cyber-eval runs. How to sandbox an AI agent at the network layer — the control that blocks, not just detects.
409,000 approve/deny decisions show humans miss 1 in 3 agent threats, and 52.5% of npm-shaped exfiltration. The AI agent permissions config that works instead.