Guide · Checked against vendor sites, September 2026
AI agent security and EDR: how they work together
The verdict
EDR and AI agent security do different jobs, and most organizations will run both. EDR detects malware and intrusion at the process level; AI agent security decides whether a specific action by a trusted AI agent should run. Several tools in this guide, including Bay and Noma Security, deploy through the EDR or MDM you already have.
Part of the University: Running it track
What does EDR see when an AI agent acts?
EDR sees processes, files and network connections. When a coding agent runs a shell command, the EDR sees a trusted, signed application doing something ordinary. Bay's research write-up on indirect prompt injection makes the point that in such attacks EDR "sees a trusted coding agent or no local process at all". The question an agent control tool answers is different: should this agent, started by this person, in this session, run this command against this data?
Source: bay.io/blog/ghostjacking-ai-agent-attack · Reviewed Sep 2026
How do AI agent security tools relate to EDR?
| Tool | Relationship to EDR (from public pages) |
|---|---|
| Bay | Deploys through your existing EDR or MDM as an ephemeral binary. |
| Noma Security | Discovers endpoint agents through existing EDR or MDM, no new endpoint agent. |
| Koi (Palo Alto Networks) | Offered as a Cortex XDR module, or standalone. |
| Prompt Security | Built into the SentinelOne Singularity Platform. |
| Onyx Security | Integrates with CrowdStrike among other security tools. |
| Lasso Security | Works with existing stacks such as Palo Alto Networks or CrowdStrike. |
| Harmonic Security | Rolls out through Intune, JAMF, Kandji or Group Policy. |
Will EDR vendors add these capabilities?
Two already have, by acquisition: Palo Alto Networks with Koi and SentinelOne with Prompt Security. For buyers on other EDR platforms, a separate agent control layer that deploys through the existing EDR is the practical route today.
What should you ask an EDR vendor about AI agents?
- Can you list every AI agent and MCP server on each device?
- Can you evaluate an agent's tool call before it runs, and ask the user?
- Can you tell an agent's action from the person's in the record?
Frequently asked questions
Can an AI agent security tool replace EDR?
No. None of the tools in this guide claims to replace malware and intrusion protection. They add control over AI agent actions.
Related
Sources
- bay.io/blog/ghostjacking-ai-agent-attack · Reviewed Sep 2026