Topic starter
06/09/2026 6:21 pm
TL;DR: Security for AI and AI for Security should be run as separate programmes that share an evidence bus, with policy at the boundary, schema-first outputs, provenance for retrieval, and continuous evaluation in CI and on shadow traffic, according to LEVO. The core shift is from model-call experimentation to routed systems with auditable controls, because the highest risk now sits where model text turns into tool action.
Editorial analysis by NHI Mgmt Group, based on content published by LEVO: “AI Security Business Narrative”.
Key questions
Q: What breaks when AI agents connect directly to tools without a gateway?
A: Direct connections create fragmented secrets, inconsistent policies, and limited visibility into what the agent actually did.
Q: Why do routed AI systems need provenance for retrieval and training data?
A: Because trust in the answer depends on trust in the sources.
Q: How can organisations prove their AI controls are actually working?
A: Look for evidence that policy decisions are logged, sensitive prompts are being redacted or blocked when required, and approved AI interactions are traceable by identity and business context.
Practitioner guidance
- Implement gateway enforcement for AI actions Route every model or agent call through a gateway that can apply input and output policy, schema validation, budgets, approvals, and trace export before any effectful tool runs.
- Adopt schema-first outputs for all AI paths Define typed outputs for free text, retrieval, and tool calls so malformed responses are blocked before they reach downstream systems or automation.
- Require provenance for retrieved and trained content Sign corpora and indexes, attach source IDs to retrieval results, and maintain takedown workflows so teams can prove where model inputs came from and whether they remain authorised.
Bottom line: AI security now depends on controlling the boundary where model output becomes tool action, because that is where most operational risk appears.
Explore further
View Full Forum → | NHI Foundation Course → | Our Services → | Read the full analysis →
06/09/2026 6:22 pm
Gateway policy is the new control plane for AI security: once model output can trigger tools, the boundary becomes more important than the model. The article correctly shifts attention away from isolated prompt security and toward policy enforcement, schema validation, budgets, and approvals at the point of action. That is the right architectural move for organisations running routed AI workflows. Practitioners should govern AI where text becomes effect, not only where text is generated.
A few things that frame the scale:
- Only 23% of IT leaders were very confident in their organisation's ability to manage security and governance for GenAI deployments, according to a 2025 Gartner survey of 360 IT leaders.
A question worth separating out:
Q: What should teams do when high-impact AI actions need approval and traceability?
A: They should keep the approval logic, tool scope, and evidence trail in one governed path so decisions are reproducible. High-impact actions need explicit human-in-the-loop criteria, bounded credentials, and exportable traces that show who approved what, what the model saw, and what action was taken.
👉 Read our full editorial: AI security needs gateway policy, provenance, and continuous assurance