“On one hand, we shouldn’t be surprised that increasingly capable agents are finding creative and sometimes unexpected ways to accomplish their objectives,” Matt Hartman, chief strategy officer at Merlin Group, told The Register.
Matt Hartman on governing agent behavior
Hartman, who joined Merlin after “a lengthy career in federal cybersecurity, including senior roles at the US Cybersecurity and Infrastructure Security Agency,” told The Register that investors are watching a specific security layer: “identity for non-human actors, clear limits on what they can access and do, and an audit trail for actions taken on an agency’s behalf.” He framed the problem as practical and immediate. Agencies, he said, are asking not only how to adopt agents but “how to constrain them and prove what one did at 2 AM on a Tuesday.”
Todd Graham on the repeating pattern of delayed security
Todd Graham, managing partner at Microsoft’s M12 venture fund, traced the dynamic to a longer pattern: new infrastructure is adopted quickly while security follows. “Every time we've built a new piece of infrastructure, we've conveniently forgotten the security,” Graham told The Register. He used familiar vendor analogies to underline the market opportunity — if laptops, cloud, or identity had been secure by default, firms such as CrowdStrike, Wiz, and Okta would not have found businesses to serve — and argued that agentic AI will be similar: “a lot of ships are going to rise with this tide.”
This site is the portfolio.
OSINTSights runs on Cloudflare Workers, D1, R2, and Vectorize, with an AI pipeline on Hetzner ARM. Nubivance designed, built, and operates it. We do the same for clients.
See what we buildAgentic identity and the old problem of service accounts
The source material ties the urgency for agent identity to a longstanding vulnerability: service accounts. These “non-human accounts” already “remain a prime target for hackers because they typically have high privileges and passwords that never expire,” the piece notes. Hartman and Graham emphasize that agents amplify an existing gap — organizations still struggle to secure service accounts even before agents began interacting with sensitive data and production systems.
AI endpoint security: “CrowdStrike for AI”
Graham described AI endpoint security as another ripe area for startups — “think of this as CrowdStrike for AI” — and suggested it is an arena he would tackle as a founder if not for a ban on starting more companies. He warned of real consequences for organizations that fail to adapt: if a breached organization is questioned for lacking antivirus or EDR, “you’re going to add AI endpoint pretty quickly to that list.” He also observed that existing endpoint and antivirus vendors “will absolutely” build products to fill the void, but sees “a moment in time where disruption is coming for everyone.”
What this means for agencies, CISOs, and startup founders
- Agencies and regulated customers: They will push for identity, access limits, and auditable trails for agent behavior — concrete governance controls Hartman says buyers are asking for now.
- CISOs at large enterprises: Graham warned they will not want to stitch together many point solutions; they will look for unified stacks that mirror human identity governance (governance, access control, authorization) but for non-human actors.
- Startup founders and investors: Both see a large market but face rising differentiation pressure. Hartman noted that AI lowers the cost to build products, which raises the bar on distinctive capabilities and go-to-market execution; Graham said he sees many founders “thinking way too small.”
Conclusion
The Register’s reporting presents a concise, repeated refrain from investors and venture executives: agents are advancing rapidly into business-critical roles, and security has lagged. The response they propose is twofold — immediate engineering of controls specific to non-human identities and rapid commercial consolidation toward platforms that provide governance, authorization and auditing for agents — and the prize is large, equated in the coverage to a "next Okta" or a “CrowdStrike for AI.” Whether startups or incumbent vendors will deliver those unified, enterprise-ready solutions fast enough is the question left standing as investors and agencies press for concrete limits on what agents can access and who can prove what they did in the middle of the night.
Original story: The Register
