AI Agents Are Becoming the Enterprise’s Newest Identities — and Security’s Next Consolidation Battleground

For decades, enterprise security has been built around a simple question: who is this user, and what should they be allowed to do? The rapid spread of artificial-intelligence agents is forcing companies to ask that question about something that is not a person at all.

As AI agents gain access to enterprise data, systems and tools, they are emerging as a new class of active identity requiring specialised permissions, monitoring and governance, according to an analysis published by Crunchbase News on September 23 by guest contributor Itay Sagie, a strategic adviser to technology companies and investors specialising in strategy, growth and mergers and acquisitions.

Sagie’s piece maps the emerging mergers-and-acquisitions landscape for AI agent security — a sign that investors and strategic buyers increasingly see this as a distinct and valuable market rather than a niche feature of existing products.

Why agents are different

AI agents differ from traditional software in important ways. A conventional application follows predetermined logic. An agent can interpret goals, plan steps and take actions across multiple systems — reading email, querying databases, calling application programming interfaces and triggering workflows.

To do that, agents need credentials and permissions. They log into systems, access sensitive data and execute transactions, often on behalf of a human user or a business process. In security terms, that makes them identities: entities with access rights that must be authenticated, authorised, monitored and, when necessary, revoked.

The challenge is that most identity systems were designed for humans and, more recently, for relatively static machine identities such as service accounts. Agents behave differently. They may be created and retired quickly, act across many systems in rapid succession and change behaviour depending on the instructions and data they receive.

The adoption gap

Industry surveys suggest that adoption has outpaced governance. Security Boulevard reported this month that 91% of organisations use AI agents, but only 10% have a developed strategy for non-human identities. The same report cited a 2026 Dark Reading poll in which 48% of security professionals ranked agentic AI as the year’s top attack vector.

Okta has reported that 81% of chief information security officers worry about excessive AI access, while only 31% feel fully aligned with their boards on acceptable AI risk, according to Security Boulevard.

That gap creates risk. An agent with excessive permissions could be manipulated through malicious instructions hidden in the data it processes, a technique often described as prompt injection. A compromised agent could exfiltrate data, alter records or initiate transactions at machine speed, potentially before a human notices anything is wrong.

Auditability is another concern. If logs cannot distinguish between actions taken by an agent and actions taken by the human it represents, investigating incidents and assigning accountability becomes far harder.

Security vendors race to respond

The largest cybersecurity companies have moved quickly to address the problem. CrowdStrike introduced its Agentic Identity Provider on September 2, 2026, positioning its identity-security platform as the control plane for what it calls the agentic enterprise. The product is designed to establish every agent as a trusted identity, grant only the access needed for as long as needed and tie every action back to the human or system behind it, according to the company.

Earlier, in June, CrowdStrike announced Continuous Identity for AI Agents, a capability that authorises agent actions in real time based on who owns the agent, who is calling it and the risk posture of their device.

Okta has also made AI agents central to its strategy. The company said Okta for AI Agents would become generally available on April 30, 2026, and its Oktane 2026 conference this month again emphasised governing agents, limiting their access and maintaining accountability, according to SiliconANGLE. Palo Alto Networks markets agentic identity-security capabilities under its Idira brand, focused on discovering, controlling and governing agent identities.

Why M&A is likely

The flurry of product launches points to why consolidation is widely expected. Large platforms want comprehensive offerings that cover human, machine and agent identities in one place, while specialist startups are building focused tools for discovering agents, managing their permissions, monitoring their behaviour and enforcing policy.

Acquisitions allow large vendors to add capabilities quickly in a fast-moving market. For startups, the prospect of acquisition by a major platform offers an attractive exit in a sector where competing against established players on distribution can be difficult.

Sagie’s analysis for Crunchbase frames AI agent security as an emerging M&A map — suggesting that the landscape of buyers and targets is taking shape as enterprises move from experimenting with agents to deploying them in production.

What enterprises should do now

For companies deploying AI agents, the practical implications are clear even as the vendor landscape evolves. Security specialists broadly recommend several principles: inventory every agent operating in the environment; give each agent a distinct identity rather than sharing human credentials; apply least-privilege access so agents receive only the permissions they need; use time-limited credentials; and maintain audit trails that clearly distinguish agent actions from human ones.

Human oversight remains important for consequential actions. Many organisations are adopting approaches in which agents can gather information and propose actions, but a person must approve high-risk steps such as moving money, changing access rights or deleting data.

A market with global implications

The shift matters well beyond Silicon Valley. Indian IT services companies, global capability centres and startups are building and deploying AI agents for clients around the world, and they will be expected to meet rising standards for agent governance and security. That creates both a compliance challenge and a business opportunity for firms that can offer secure agent deployment as a service.

Indian cybersecurity startups, too, may find openings in specialised areas such as agent discovery, policy management or monitoring — segments where focused innovation can compete with larger platforms.

The bottom line

AI agents promise significant productivity gains, but they also introduce a new kind of risk: autonomous software with real access to real systems. Treating agents as identities — with the same rigour applied to employees and critical systems — is quickly becoming a baseline expectation.

As vendors race to build and buy the tools needed to manage that risk, agent identity security is shaping up to be one of the most important and competitive segments of the cybersecurity market in the AI era.