Topic starter
25/10/2026 11:00 am
TL;DR: Gartner says 80% of unauthorized AI agent transactions will stem from internal policy violations by 2028, according to DTEX. Agentic workflows need runtime context and decision trails to stay reviewable.
Editorial analysis by NHI Mgmt Group, based on content published by DTEX: “Gartner® Market Guide for Guardian Agents”.
By the numbers:
- At least 80% of unauthorized AI agent transactions through 2028 will come from internal policy violations rather than malicious attacks, according to Gartner research cited by DTEX.
Key questions
Q: Why do autonomous agents create new governance risks in enterprise workflows?
A: Autonomous agents complicate governance because they can act across systems, carry context between steps, and make decisions without a human at each hop.
Q: What are the signs that an AI agent session is drifting out of policy?
A: Look for unexpected tool calls, suspicious command sequences, credential exposure attempts, memory manipulation, or changes that exceed the user's stated objective.
Q: What breaks when AI agents can make code changes faster than humans can review them?
A: Manual review stops being a meaningful control if it cannot keep pace with change volume.
Practitioner guidance
- Instrument agent activity with behavioural context Capture the sequence of tools, files, pages, and data movements around each agent action so analysts can see what happened before, during, and after a decision.
- Separate acting agents from reviewing agents Design a control path where one component performs the task and another evaluates the outcome against policy and confidence thresholds before closure.
- Map agent permissions to task scope Review whether each agent has only the access needed for the specific workflow it performs, and remove standing access that outlives the task.
Bottom line: AI agents create a governance gap when they act at machine speed inside workflows built for human review.
Glossary terms: Guardian agent, Behavior Baseline, Decision Governance, Agentic workflow
Explore further
View Full Forum → | NHI Foundation Course → | Our Services → | Read the full analysis →
This topic was modified 7 days ago by NHI Mgmt Group
27/10/2026 7:14 pm
Guardian agents are becoming a governance layer, not just a detection feature. The article makes clear that enterprises need a second control plane for AI systems that can act at machine speed. That shifts the question from whether an agent behaved strangely to whether its action was acceptable before execution completed. For IAM and PAM teams, the implication is that runtime evaluation must sit alongside identity issuance and entitlement review.
A few things that frame the scale:
- 53% of security leaders expect AI to run major portions of their infrastructure autonomously within the next three years, according to the 2026 Infrastructure Identity Survey.
A question worth separating out:
A: They need a model where agents can investigate and propose actions, but humans still own the judgment calls and the verification standards. The practical guardrail is a system that shows what the agent did, what remains pending approval, and whether the action actually worked. That preserves accountability while reducing manual toil.
👉 Read our full editorial: Guardian agents expose the oversight gap in agentic AI workflows
This post was modified 7 days ago by NHI Mgmt Group