Machine identities outnumber humans 80 to 1, and four vendors are bidding to own the gap

The figure is 80 to 1. Machine identities now outnumber human identities in the enterprise by more than that margin, according to Palo Alto Networks, which published the ratio in the release confirming it had completed its acquisition of CyberArk on February 11, 2026 and made Identity Security a core pillar of its platform strategy ^[1][2]. CyberArk shareholders receive $45.00 in cash and 2.2005 Palo Alto shares per ordinary share ^[3].

What the ratio turns on is standing privilege. Palo Alto says 75% of organizations acknowledge their human identities are governed by outdated, overly permissive privilege models, that nearly 90% have already suffered an identity-centric breach, and that identity-driven controls can cut breach response time by up to 80% ^[4][5][6]. Apply a permissive model to a population 80 times larger that runs continuously, and the review cycle that just about worked for humans stops working. CrowdStrike's George Kurtz puts the same point more bluntly: every agent is a privileged identity ^[13].

Hence the buying. CrowdStrike signed for SGNL on January 8, 2026, acquiring the runtime enforcement layer that sits between identity providers and the SaaS and cloud resources identities actually reach, and extending just-in-time access beyond Active Directory and Entra ID to AWS IAM and Okta ^[8][9][10]. Cisco is folding WideField into Splunk to correlate identity, session and activity telemetry, following Astrix and Galileo, and frames identity as who or what took an action, under which authority, in which session, with what blast radius ^[14][16][15]. Microsoft, meanwhile, calls Agent 365 the control plane, with an Entra registry as single source of truth, quarantine for unsanctioned agents, and shadow agents promised later; Foundry registers each agent there automatically ^[17][19]. IDC's forecast of 1.3 billion agents by 2028 is the demand case ^[18].

The contest is visible in Microsoft's own documentation. Blocking an agent in the Microsoft 365 or Teams admin center only affects its projection into Teams and Copilot; the agent stays fully functional in Foundry and other integration points, with no impact on underlying Azure resources ^[20]. And an AI Administrator cannot elevate their own access to intervene, and must go through the agent owner or a Global Administrator ^[21]. The registry knows the agent. The administrator cannot always stop it.