AI security summits in 2026 reflect rising agent governance risk

At a glance

What this is: Akto’s 2026 summit roundup shows AI security events centring on agents, MCP workflows, threat detection, and governance rather than generic model talk.

Why it matters: That matters because practitioners are being pushed to treat AI agents as governed systems with access, accountability, and control boundaries, not just productivity features.

Context

AI security conferences are becoming a proxy for where the market thinks control gaps are opening up. In this case, the article is less about event logistics than about the governance problems now surrounding AI agents, MCP workflows, shadow AI, and operational security posture.

The identity angle is real: once agents can act, call tools, and move through workflows, security teams have to decide how those systems are authenticated, authorised, monitored, and retired. That makes the topic relevant to IAM, PAM, NHI governance, and emerging AI agent identity programmes.

Key questions

Q: How should security teams govern AI agents that can choose tools at runtime?

A: Security teams should govern runtime agent choice as an access event, not as a simple application action. That means scoping permissions to the task, limiting token lifetime, logging every tool decision, and blocking the agent from reaching systems outside its approved context. Static roles alone are not enough when the execution path changes on each run.

Q: Why do AI agents and MCP-connected workflows create new data exposure risks?

A: AI agents and MCP-connected workflows expand access beyond human users into systems that can retrieve, transform, and move sensitive data automatically. That increases the chance of uncontrolled sharing, especially when agents can reach code repositories, tickets, cloud storage, and internal knowledge bases. Security teams need policy enforcement, access visibility, and content inspection around every agent action.

Q: What are the warning signs that AI governance is becoming a control gap?

A: The warning signs are fragmented inventories, undocumented OAuth grants, unclear ownership of AI tools, and inconsistent evidence for what each system can access. If business teams can deploy AI faster than security can enumerate and review it, governance is already lagging. That is the point where the programme needs cross-platform controls, not more isolated approvals.

Q: Should organisations treat AI agents like human users in IAM?

A: No. Human IAM assumes a person logs in, works within a session, and can be reviewed later as a stable identity holder. Agents can act at machine speed, across multiple systems, and with changing runtime context, so they need identity governance built around execution and delegation rather than human authentication patterns.

Technical breakdown

Why AI agents create a governance problem, not just a model problem

AI agents change the control surface because they can execute actions, invoke tools, and chain decisions in ways that exceed a simple model prompt-response pattern. That means the security question is no longer only whether the model is safe, but whether the agent’s permissions, tool scope, and runtime behaviour are governed. In practice, an agent becomes a software actor with identity-like properties: it needs issuance, authorisation, revocation, and logging. That creates a direct overlap with NHI governance, because unmanaged agents behave like privileged non-human identities even when they are presented as application features.

Practical implication: define AI agents as governed identities with explicit lifecycle controls, not as informal automation.

MCP workflows expand the attack surface for access and delegation

MCP, the Model Context Protocol, connects agents to tools and data sources, which makes delegation and trust boundaries much more important than the model itself. The protocol does not remove risk, it concentrates it at the point where an agent is allowed to reach into systems, request context, and act on it. That is why governance needs to focus on what the agent can access, how consent is granted, and whether delegated actions are bounded to a task, session, or business purpose. Without that, MCP becomes a convenient route for overbroad tool use and ambiguous accountability.

Practical implication: inventory every MCP-connected tool and constrain delegation to least-privilege task scope.

Shadow AI is really shadow governance

Shadow AI is not only an inventory issue. It is what happens when teams deploy or connect AI systems faster than the organisation can assign ownership, review access, or define acceptable use. In that sense, the article points to a broader governance failure: if security leaders cannot see which agents exist, what they can do, and who is responsible for them, then controls cannot be enforced consistently. That problem maps closely to the way unmanaged service accounts and orphaned credentials create hidden risk in traditional identity programmes.

Practical implication: establish ownership, discovery, and offboarding controls for every AI system before broad deployment.

Threat narrative

Attacker objective: The objective is to exploit agent trust and delegated access to expand control over data, workflows, or security decisions without adequate oversight.

1. Entry occurs when AI agents are connected to enterprise tools or MCP workflows without a clear governance boundary around their access.
2. Escalation follows when those agents inherit broader permissions than the task requires, allowing them to reach data or actions beyond intended scope.
3. Impact emerges when shadow AI or poorly governed agents create unauthorised access paths, weak accountability, or operational decisions that security teams cannot reliably trace.

Breaches seen in the wild

• CoPhish OAuth phishing via Copilot Studio: Datadog showed Copilot Studio agents on a Microsoft domain can front OAuth consent phishing and forward stolen tokens; no victims reported.
• 12,000 secrets in LLM training data: Truffle Security found 11,908 live API keys and passwords hard-coded in web pages captured by Common Crawl, a dataset used to train LLMs.

Read and download The State of NHI & AI Agent Breach Report 2026, covering 200+ breaches impacting Non-Human Identities including AI Agents.

NHI Mgmt Group analysis

AI security summits now function as governance signal events. The conference agenda is shifting toward AI agents, MCP, shadow AI, and threat detection because those are the places where control failures are becoming visible. That means the market is moving from AI experimentation to AI supervision, and practitioners should read the agenda as a map of emerging control debt.

Agent identity is becoming a first-class security problem. When an AI system can request tools, hold context, and initiate actions, it behaves like a non-human actor that needs lifecycle management. The governance model for that actor cannot be borrowed wholesale from human IAM, because the agent’s access may be dynamic, transient, and embedded in workflow execution.

AI governance debt: the backlog of controls created when agent deployment outruns ownership, authorisation, and offboarding. This is the failure mode the article points to most clearly, even if it does not name it directly. Once teams cannot say who owns an agent, what it may touch, and when it should be retired, security posture degrades faster than model capability improves.

MCP makes delegation the real control point. The protocol itself is not the problem, but it forces teams to confront tool scope, context sharing, and approval boundaries. For IAM and PAM teams, this means access policy must move closer to runtime behaviour, because static registration alone will not describe what an agent can actually do.

Summit hype only matters if it changes operating discipline. Conferences can surface ideas, but practitioners need to translate them into policy, inventory, and review processes. The useful outcome here is not curiosity about AI security as a category, but clearer ownership of AI systems as governed assets.

From our research library:

• Only 13% of organisations feel extremely prepared for the reality of agentic AI despite the majority racing toward autonomous adoption, according to the 2026 Infrastructure Identity Survey.
• Gartner predicts that by the end of 2026, 40% of enterprise apps will feature task-specific AI agents.
• Read next: Agentic AI Identity Maturity Model

What this signals

Agent governance will become a board-level control topic faster than many teams expect. Once AI systems can initiate work across tools and workflows, security leaders need a way to answer the basic questions of ownership, authority, and revocation. That shifts agent security from experimentation to governance, and it brings IAM, PAM, and NHI disciplines into the same conversation.

Identity for AI systems is becoming an operating model issue, not a naming exercise. The hard part is not calling something an agent, but deciding what it is allowed to do, how long that permission lasts, and who can disable it. Teams that leave those questions unresolved will end up with policy statements that do not match runtime behaviour.

Only 13% of organisations feel extremely prepared for the reality of agentic AI despite the majority racing toward autonomous adoption, according to the 2026 Infrastructure Identity Survey. That gap suggests governance lag, not technology maturity, will be the first operational constraint for most programmes.

For practitioners

• Define an AI agent inventory Track each agent, its owner, its tool connections, and the business purpose it is allowed to support. Include unmanaged prototypes and pilots so shadow AI cannot accumulate outside policy.
• Bind MCP access to task scope Limit each tool connection to the smallest required dataset, action set, and session duration. Review delegation rules whenever an agent gains a new tool or workflow.
• Treat agent offboarding as a control Remove credentials, API access, tokens, and workflow hooks when an agent is retired or replaced. Offboarding must be explicit because stale agent access becomes hidden standing privilege.
• Separate experimentation from production Keep proof-of-concept agents out of production systems until ownership, logging, and approval boundaries are in place. A pilot that can touch live systems is already a governance decision.

Key takeaways

• AI security summits in 2026 are reflecting a shift from general model risk to the governance of agents, tools, and delegated access.
• The central control issue is who owns an AI agent, what it can touch, and how its authority is revoked when it is no longer needed.
• IAM and NHI teams need to treat AI agents as governed actors with lifecycle controls, because runtime behaviour now matters as much as model quality.

Key terms

• Agentic AI: Autonomous AI systems capable of planning, deciding, and taking actions, including calling APIs, writing code, and orchestrating other agents, with minimal human oversight. Agentic AI introduces new NHI risks as agents must authenticate to external services.
• MCP Workflow: An MCP workflow is the sequence of steps an AI agent follows when using the Model Context Protocol to request tools, data, or actions. It defines how the agent discovers capabilities, sends structured requests, receives responses, and continues execution. In security terms, it creates a governed path for agent-to-system interaction.
• Shadow AI: AI agents, copilots, or connected tools operating without full visibility or governance from security teams. Shadow AI becomes an identity problem when those systems authenticate with unmanaged tokens, service accounts, or OAuth apps that can reach production resources.
• Agent Identity: An agent identity is the set of attributes, credentials and permissions assigned to an autonomous software entity. It is treated as a non-human identity because it can authenticate, act on systems and accumulate access over time, which creates governance, audit and lifecycle obligations similar to other production identities.

Deepen your knowledge

NHI Foundation Level course, the industry's only accredited NHI security programme, covers NHI governance, machine identity security, and agentic AI identity. It helps security and identity practitioners translate emerging agent risk into operational controls.