AI Security

Secure60 AI Security covers the AI applications an organisation runs, protects and uses without approval. It is built on the same context as the rest of the platform: AI calls are events in your project, AI applications and people are entities, and AI detections are rules that raise threats in the same queue as every other threat.

Four parts work together:

Part	What it does	Where
AI Activity	One page listing every AI application in use, who uses it, which model provider it reaches, and the full audit trail of sessions, calls and guardrail findings	AI → AI Activity
Secure60 AI Protector	A gateway in front of the model APIs your applications call. It checks every prompt and reply for personal data, secrets and prompt injection, masks or blocks by policy, and records the call	Integrations → Secure60 Protector
Discovery	AI applications the Protector serves, and AI services seen in proxy, DNS and firewall logs, appear on AI Activity as Discovered until someone registers them	AI Activity, collector parser templates
Lifecycle and governance rules	Each AI application carries a state (Discovered, Under review, Sanctioned, Restricted, Prohibited, Retired) with the reason recorded; managed rules raise threats when a Prohibited or Retired application is used	AI Activity, Detection Rules

Documentation

• AI Activity — The Overview, AI Audit Trail and Watch tabs: KPIs, the who-uses-what flow chart, sources, applications and the session view.
• Discover AI Applications — How the Protector’s seen list and the AI Traffic Detection parser template find AI applications, and what each source can see.
• AI Application Lifecycle — The six states, what each one does for detection, recording a reason, and the AI Governance rule pack.
• AI Audit Trail — What is recorded for each AI call, how sessions are built, search and retention.
• Prompt-Attack Detection — Prompt-injection detection in the AI Protector and the managed rules that raise threats from it.
• LLM Data Exfiltration — Personal data and secrets in prompts, tool results, replies and tool calls: detect, mask, block and alert.
• AI Governance Frameworks — ISO 42001, NIST AI RMF and EU AI Act templates in the Governance capability.
• Secure60 AI Protector — Deploying and configuring the gateway: backends, guardrail policies, identity, capture modes.

Where each source fits

AI Activity combines four sources. Each one sees a different amount of an AI interaction:

Source	Sees	Does not see
Secure60 AI Protector	The prompt, the new messages of each call (including tool results), the reply, the guardrail decision and every finding; model, token counts and tool names	Calls that do not pass through it
Secure60 agents	Every step, tool call and decision of a Secure60 agent session	— (model and token counts of Secure60 agents are not shown)
Vendor audit logs	Records sent with operation = ai-usage: who, which application, which files	Prompt and response text
Network and endpoint	AI services reached from proxy, DNS and firewall logs tagged by the AI Traffic Detection parser template: destination, host, user, volume	Content, model and purpose

The Protector is the only source that records content. An application seen only on the network is listed with Network only visibility; routing it through a Protector adds the prompt, reply and findings.

How it fits the platform

AI Security capability	Built on
Prompt-injection, personal data and secret detection on AI calls	Secure60 AI Protector
AI audit trail	Log Management — Protector and agent events in your project
Shadow AI discovery	Entity tracking from the Protector and the collector, plus the AI Traffic Detection parser template
AI application register and lifecycle	Entities (type app) with a lifecycle state
AI threats	Threat Detection — the AI Protector and AI Governance managed rule packs
Teams in AI charts	Entity groups (users in a group)
AI governance frameworks	Governance capability

Getting started

1. Deploy a Secure60 AI Protector in front of one application, with its policy in Monitor mode.
2. Link the AI Traffic Detection parser template to the collector profile that receives your proxy or DNS logs (Discover AI Applications).
3. Deploy the Secure60 – Managed Rules – AI Protector and Secure60 – Managed Rules – AI Governance packs to the project under Detection Rules.
4. Open AI → AI Activity, review the applications listed as Discovered, and set each one’s state (AI Application Lifecycle).