Agentic AI explosion driven by $50B market opportunity and related risks

The term “agentic AI” is now entrenched in the digital lexicon, used to refer to the category of machine learning agents that are assigned various tasks in a digital ecosystem. A lot of the chatter about these worker bots is about how they’ll transform business – but just as much is about how to know if we can trust them to do what they’re told.

The agentic AI pitch can be seen through two lenses, the first economic, the second more about psychology. A recent research report projects that the market will grow from around 7.8 billion dollars in 2025 to over 50 billion by 2030. That said, the amount that Big Tech has invested in infrastructure for this tech means it has money to recoup. Moreover, as large language models and machine learning tools become more common, employees are turning to unmanaged AI tools that make their lives easier, even if they aren’t approved by IT.

The race is on to harness the potential in AI agents, both by deploying them, and taming them.

HID says time has come for evolution of PKI

Citing Gartner, a blog from HID says that by 2026, over 40 percent of enterprise workflows will involve autonomous agents such as AI copilots, workflow bots or digital assistants. “Already, organizations are deploying AI-driven assistants to generate code, write reports and manage supply chains.”

According to AWS, more than half of enterprises now see ‘agentic AI’ as a strategic priority for digital transformation. Yet HID asserts that most enterprises still lack the identity and access management (IAM) frameworks needed to govern these machine learning models don’t go off-script. “Consider a rogue ‘procurement agent’ impersonating a legitimate workflow bot to issue unauthorized purchase orders within a supply chain system,” HID says. “Or a rogue assistant accessing confidential data through a compromised integration. Without verifiable identity and authorization, these scenarios quickly move from hypothetical to inevitable.”

The time has come for a global trust model for AI, and the evolution of Public Key Infrastructure (PKI).

HID ANS functions as PKI-backed directory for agentic governance

HID believes a promising approach is to be found in the Agent Name Service (ANS), a draft standard proposal for the Internet Engineering Task Force (IETF).

“Just as DNS maps human-readable names to IP addresses, ANS maps agent identities to verified capabilities, cryptographic keys and endpoints,” it says. “It functions as a universal, PKI-backed directory for secure agent discovery and communication.”

Agentic machine learning systems can spawn, evolve and retire agents quickly, so capability attestation – what it’s allowed to do – is of critical importance. Operating across diverse ecosystems such as Model Context Protocol (MCP), Agent to Agent (A2A) and Agent Communication Protocol (ACP) requires interoperability of trust. Fifteen percent of organizations have already begun deploying certificates for AI agents, according to the HID PKI Market Study.

As the agentic multitude grows, so grows the attack surface for sybil attacks, tampering, man-in-the-middle impersonation and other threats. HID says ANS “addresses these risks by applying cryptographic integrity and life cycle governance to every agent interaction. Its model mirrors DNS governance under ICANN, ensuring name uniqueness and trust transparency through federated authorities.”

“Long-term success will depend not only on cryptography, but also on policy, oversight and auditability – extending governance frameworks such as ICAM and eIDAS into the world of autonomous AI.”

HID PKIaaS certified compliant with TX-RAMP requirements

HID has also announced that its HID Public Key Infrastructure as a Service (PKIaaS) is now officially certified by the Texas Risk and Authorization Management Program (TX-RAMP). A blog post calls it “a significant achievement that reinforces our commitment to delivering secure, compliant and reliable cloud-based PKI Services and Certificate Lifecycle Management (CLM) to the public sector in Texas.”

TX-RAMP standardizes the assessment of security of cloud services used by Texas state agencies and institutions of higher education. Certification, for which vendors must meet NIST 800-53-based security controls, is mandatory for any cloud service provider looking to do business with Texas public sector entities.

HID says the certification “simplifies procurement for agencies bound by TX-RAMP requirements, strengthens trust in HID’s ability to safeguard sensitive data and digital identities, and supports secure integrations with over a dozen enterprise tools.”

Mandate Framework from authID binds agents, humans with biometrics

An announcement from authID introduces the Mandate Framework, a “comprehensive governance model for agentic AI security with support for non-human identities, including autonomous and semi-autonomous AI agents.”

The company says the present crop of agents operates with weak, static, phishable credentials and tokens that did not originate with biometric and cryptographic binding to a verified human. Conversely, with Mandate, each agent is “sponsored” by a verified human to operate within authorized boundaries, and the platform produces auditable immutable records.

“Enterprises are ready to let AI agents work, but not without a governance framework,” says Rhon Daguro, CEO of authID. “The authID Mandate Framework is our blueprint for accountable agentic AI. Customers can govern which agents take action, who sponsors them, and what they are allowed to do, before, during, and after every action.”

Erick Soto, chief product officer at authID, points out that “just last week, fraudsters successfully orchestrated attacks on over 30 companies through AI agents. This incident shows how quickly AI tools can be weaponized when there’s no clear chain of trust.” He says that, by operationalizing issuance of biometric-rooted, cryptographically verifiable credentials from human sponsors and enforcing them across digital workflows, the Mandate Framework “provides the assurance these CEOs need to launch their AI solutions with confidence.”

Microsoft gazes into the agentic AI frontier

If you want proof machine learning agents are going mainstream, simply look to Microsoft.

At this year’s Ignite conference, AI and its agents made headlines, notably for powering new capabilities in Microsoft 365 Copilot, and in the form of Microsoft Agent 365, a new “control plane” that functions as a central management dashboard within the Microsoft 365 Admin Center.

The intention behind Agent 365 is to give IT departments the equivalent of an HR department for bots, by assigning every agent, from any company, a unique identifier, allowing companies to use existing security systems to manage and control agent interactions. Microsoft to the table brings the advantage of its ubiquity in corporate identity and security systems; “instead of asking companies to adopt an entirely new platform, it’s building AI agents into tools that many businesses already use.”

Of the Silicon Valley titans, that makes Microsoft, Google, Amazon and Salesforce that have now rolled out agentic platforms for corporate use.

A post from Micrtosoft’s blog lays out a vision for the future of work shaped by so-called Frontier Firms – “organizations that are human-led and agent-operated.” As such, it has launched the Frontier Firm AI Initiative with the Digital Data Design Institute at Harvard, which aims to “research human-AI collaboration and equip global C-suite leaders with the skills to transform their organizations into Frontier Firms.” The first cohort includes big names such as Barclays, Eli Lilly and Company, Mastercard, Levi Strauss, Lumen and Nestlé.

EMVCo encourages stakeholders to join discussions on agentic specification

EMVCo, the technical body that creates and manages specifications for the EVM payments system, has announced that it is exploring a global specification to support agentic payment solutions. A statement from the organization says that, “as industry adoption and innovation accelerate, a globally interoperable and scalable approach may be beneficial in realizing trusted agentic payments for consumers, merchants and issuers.”

As such, EMVCo is working on how EMV Specifications, including EMV 3-D Secure (3DS), EMV Payment Tokenization and EMV Secure Remote Commerce (SRC), can be developed and enhanced to promote seamless and secure card-based agentic payments.

Patrik Smets, EMVCo executive committee chair, says there is “a clear opportunity to collaborate with participants across the industry to extend this experience to support agentic payment solutions.”

EMVCo is encouraging all interested stakeholders, including global technical bodies, to explore ways to participate in EMVCo and “share input on the strategic considerations and technical advances shaping agentic commerce and payments.”

Agentic commerce looks set to become the next frontier, so to speak, in the ongoing development of agentic AI. Startups like HUMAN Security are looking to capitalize on the need for guardrails on agentic shopping. Meanwhile, Amazon is suing Perplexity AI over an automated ordering feature it says covertly accessed customer accounts.

The story points to a potential snag in the great rise of agentic AI: to paraphrase a familiar proverb, one company’s agent is another company’s pest.

Article Topics

AI agents  |  authID  |  biometric binding  |  EMVCo  |  HID  |  identity access management (IAM)  |  Microsoft  |  PKI