NVIDIA: AI security is an engineering problem — defend every layer of the agent stack
NVIDIA Blog · rss · 2026-09-21
NVIDIA argues AI security must be treated as engineering: defined requirements, enforceable controls, named owners, and evidence of protection.
- Security spans the full agent stack: models, harnesses, and runtime environments each carry responsibilities; controls must follow data, instructions, and actions.
- Boundaries must hold even when an agent errs: runtime should limit files, network destinations, and processes independently of the agent's reasoning; permission to update a record shouldn't imply permission to export it.
- Every agent needs traceable identity, least-privilege credentials, human approval for consequential actions, and protected logs for forensics.
- Tools: open-source NVIDIA OpenShell (sandboxed runtime enforcing policies beyond the agent's reach), with Cisco DefenseClaw, JFrog skill scanning, CrowdStrike SafeMind, Palo Alto Prisma AIRS, Capital One VulnHunter, ReversingLabs Spectra Assure.
- Teams need pre-deployment evidence (credential theft and exfiltration blocked), reproducible failure investigation, and open sharing of what works to shift advantage to defenders.
More from coding & agent
- User automates entire video clipping workflow with Nous Research's Hermes agent — Teknium · 2026-11-12
- As models improve, writing good evals gets harder and more important — amankhan · 2026-11-12
- Free AI game jam in SF Nov 14–15 with teams from Anthropic, Google, Meshy, OpenAI — chrisfirst · 2026-11-12
- Omarchy 4.6 lets you share local AI models across your tailnet, says DHH — alexcovo_eth · 2026-11-12
- Coding harnesses hand shell access to inference providers — 4-step gateway mitigation — shensi · 2026-11-12
- Slack's David Hoang: agents beat people on coordination, proposes 'boundary agents' — davidhoang · 2026-11-12