At Black Hat USA this week, ServiceNow's cybersecurity leadership opened with an uncomfortable number: the average enterprise now runs more than 70 different security tools, and most of them were never built to talk to each other fast enough to matter. Machine identities, service accounts, API keys, AI agents, are doubling roughly every 18 months, and a growing number of them were provisioned by nobody in particular, for a purpose nobody wrote down.

Yevgeny Dibrov knows that problem from both sides. He co-founded and ran Armis, the asset visibility company ServiceNow acquired earlier this year, and now leads ServiceNow's combined cybersecurity and risk business. His framing of the gap was blunt: security teams are still operating on a human clock while the systems they're defending compound risk on a machine one. "Fragmented security tools can't match the curve AI is creating," he said.

Why security operations and risk management ended up on the same platform

Cybersecurity and enterprise risk management grew up as separate disciplines with separate tools. Security operations centers ran SIEMs, EDR platforms, and ticketing systems built for speed: detect a threat, contain it, move on. Risk and compliance teams ran GRC software built for documentation: policies, control libraries, audit trails, vendor questionnaires. The two groups reported to different executives, measured different things, and rarely shared a data model.

The problem with keeping them apart is that the underlying facts don't actually separate that cleanly. A vulnerability that never gets patched is a security problem and a risk exposure. A vendor whose posture degrades is a third-party risk and a potential incident waiting to happen. An AI agent with excessive permissions is a governance failure and an attack surface, at the same time. Treating these as three tickets in three systems, owned by three teams that don't share context, is how a known issue sits unpatched for months while everyone assumes someone else is tracking it.

ServiceNow's answer has been to group its work in this space around six areas that all draw from the same asset and identity data rather than around departmental lines: unified exposure management, continuous vulnerability detection, cyber-physical security, identity and access security, agentic incident response, and cyber risk and compliance. None of that requires a security team and a compliance team to merge organizationally. It just means they stop working from five different pictures of the same environment.

What's actually in this part of the platform today

A useful way to think about what ServiceNow sells here is three families of products that have grown up over several years, plus two recently acquired capabilities that now run through all of them.

Security operations covers the incident-facing work. Security Incident Response ingests alerts from SIEMs, EDR tools, and endpoints, adds asset and business context from the CMDB, and automates the coordination that used to require manually pulling three teams into a war room. Unified Security Exposure Management, the product long known as Vulnerability Response, consolidates findings from scanners, cloud tools, containers, and code into one workflow and prioritizes by exploitability and business impact rather than raw CVE counts. Threat Intelligence Security Center fuses internal incident data with external threat feeds so analysts hunt and model threats with full business context instead of switching between disconnected tools. Security Posture Control checks continuously whether controls like endpoint protection or firewall rules are in place across every asset, catching gaps that vulnerability scanning alone misses. Newer additions extend this further: Application Security now threat-models AI-generated code and its dependencies, Dynamic Application Security Testing validates runtime vulnerabilities in live applications and APIs, and External Attack Surface Management shows an organization's exposed footprint the way an attacker would see it.

Risk and compliance covers the documentation-facing work: Policy and Compliance Management (the control library and policy lifecycle), Risk Management (a risk register scored against real asset and incident data), Audit Management, Third-Party Risk Management, Business Continuity Management, Privacy Management, and Environmental, Social, and Governance Management for organizations tracking those obligations alongside everything else. This is also where the terms "GRC" and "IRM" get used almost interchangeably, and where a fair bit of buyer confusion lives, worth a quick note in the FAQ below. A newer capability here, Cryptographic Asset Compliance, helps organizations discover and migrate legacy cryptographic algorithms toward quantum-resistant standards, feeding evidence directly into the same compliance workflows.

Identity and access security, largely powered by the recently acquired Veza, governs who and what can actually do things across the environment, human employees, service accounts, and AI agents alike. Two capabilities sit on top of it now: AI Agent Access Security, which unifies access control for agents regardless of which model or platform built them, and Non-Human Identity Remediation, which moves past simply scoring risk into taking action, rotating keys, deprovisioning, and revoking permissions automatically at scale.

Two acquisitions behind almost everything new here

Most of what makes this part of ServiceNow look different today than it did twelve months ago comes down to two deals. In December 2025, ServiceNow announced its intent to acquire Veza, a leader in identity security, for roughly $1.3 billion; the deal closed on March 2, 2026. Weeks later, ServiceNow agreed to acquire Armis, a cyber exposure and asset visibility company, for $7.75 billion, the largest acquisition in ServiceNow's history; that deal closed on April 20, 2026.

Veza brings an Access Graph: a continuous, real-time map of exactly what every identity, human or otherwise, can do across an organization's cloud, SaaS, and on-premises systems. That distinction matters because traditional identity tools stop at directory groups and role names. Veza traces the full path from a user or service account down to the specific resource it can touch, which is the difference between knowing someone has a role and knowing what that role lets them do. It governs human and non-human identities under one framework, surfacing overprivileged or dormant accounts and enforcing least privilege at the point of action.

Armis solves a different, older problem: you can't secure what you can't see. It delivers agentless, real-time discovery across IT, OT, IoT, medical devices, code, and cloud, tracking billions of connected devices and enriching every asset record with device type, firmware version, behavioral data, and live risk posture. That intelligence flows directly into the CMDB, turning what used to be a static inventory into a live picture of the actual attack surface.

"Effective cybersecurity starts with understanding risk and maintaining visibility across the enterprise."

Brandon Glaze, Sr. Director Cybersecurity (OT/ICS), Baker Hughes

Together, Armis and Veza now feed ServiceNow's AI Control Tower and a newer orchestration layer called Context Engine with the asset and identity intelligence needed to act on risk rather than just document it. The two acquisitions were first folded into a combined offering called Autonomous Security & Risk, unveiled at ServiceNow's Knowledge conference in May 2026. This week at Black Hat, that vision moved a step further, formalized as six unified solutions under the Autonomous Security name, with cyber risk and compliance now standing as one of those six pillars rather than a separate business.

Shift Zero and the case for prevention over detection

The philosophy tying all of this together is one ServiceNow calls Shift Zero: the idea that the goal was never to detect problems faster, it was to stop them from entering the environment at all. In practice that means embedding security checks into the workflow where code gets written, assets get provisioned, and access gets granted, rather than scanning for what already went wrong afterward. The clearest example ServiceNow points to is a security agent running as a sidekick to a coding agent, validating and correcting code in real time as it's written so vulnerabilities never make it into a deployed system in the first place.

Nadir Izrael, who co-founded Armis alongside Dibrov and now leads ServiceNow's newly formed AI Center for Cyber Defense, put the shift in concrete terms: with Shift Zero, "we now prevent problems from entering the system at all," rather than finding a problem and then fixing it. The center is built around a two-sided mandate the team calls AI for security and security for AI: using AI models defensively to catch attacks that now move at machine speed, while simultaneously governing the AI agents and models an enterprise deploys so they don't become the next attack surface themselves.

Every AI agent is an identity, and most aren't governed

The reason identity security has become the fastest-growing piece of this business isn't abstract. As AI agents take on more autonomous work, each one operates under an identity, with a set of permissions almost always designed for a human user rather than a system acting continuously at machine speed. Non-human identities already outnumber human ones in most enterprise environments, and few organizations can say with confidence who approved a given agent's access, why it was granted, or whether it's still needed.

Dibrov has described where this is heading in terms borrowed from an earlier security era: in the same way zero trust became the foundational architecture for the cloud, he expects "zero permissions" to become the foundational architecture for the agentic one. That shift shows up concretely in results ServiceNow has reported from its own customer base: a major U.S. financial services institution eliminated 96% of its dormant non-human identities after deploying Veza, turning least privilege from a policy statement into something enforced.

"Every action performed by a non-human identity must be traceable to an accountable human owner to ensure legal attribution, oversight, and responsibility."

Jeffrey DiMuro, Deputy Chief Information Security Officer, ServiceNow

That's not a hypothetical concern. Independent research from IBM found that one in six security breaches in 2025 already involved AI in some form, and 97% of businesses that reported an AI-related breach lacked proper AI access controls at the time. Separately, identity provider SailPoint reports that 80% of organizations have already seen AI agents act outside their intended boundaries, including unauthorized access, handling of restricted information, and phishing-adjacent behavior.

What the new AI specialists actually do

Two of the newest pieces of this rollout are purpose-built AI specialists rather than dashboards or workflow templates. The Vulnerability Resolution AI Specialist orchestrates triage and remediation at scale, executing low-risk patches on its own and turning what used to be an ever-growing exposure backlog into something closer to a closure pipeline. The Tier 2 SOC AI Specialist autonomously builds and executes multi-phase response plans for complex incidents, handling enrichment, correlation, containment, and blocking, and escalating only the high-risk decisions to a human analyst. Both are expected to reach general availability in December 2026, alongside Agentic AI for Continuous Control Monitoring, which turns control evidence into a real-time operational signal rather than a quarterly exercise, and Cryptographic Asset Compliance for quantum-readiness work.

A number of related capabilities are already available now rather than pending: Agentic Exposure Management, Application Security, DAST, EASM, Agentic AI for Cyber-Physical Security, AI Agent Access Security, and Non-Human Identity Remediation. All of it runs under the governance of ServiceNow's AI Control Tower, which inventories agents from the moment they appear, scores their risk continuously, and enforces least privilege in real time.

What the data says about where risk is heading

ServiceNow's 2026 Risk and Security Outlook, produced with research partner ThoughtLab from a survey of 1,000 risk and security leaders across 11 countries and 11 industries, paints a fairly stark picture of the gap between rising exposure and organizational readiness. Fifty-seven percent of executives say technology and AI-driven risk has grown sharply over the past three years, and a similar share expect that trend to continue over the next twelve months. Cyber adversaries are weaponizing AI specifically to sharpen phishing and social engineering, exploit insider access, and accelerate data and IP theft, all cited by roughly a third or more of executives as areas where AI will meaningfully raise threat intensity in the coming year.

Confidence hasn't kept pace with that growth. Close to half of surveyed organizations report only very low to moderate confidence in their overall security and risk posture, and confidence dips further when the question turns specifically to AI-enabled attack vectors: 57% report low confidence in their ability to handle AI-driven cloud misconfigurations, the weakest spot identified in the entire survey. Only 52% of organizations report having taken meaningful steps to improve their risk preparedness at all over the past year, despite the acknowledged rise in exposure. Government agencies and consumer-facing industries with lighter historical investment in risk tooling report the least confidence; financial services, life sciences, and insurance, sectors that have invested more consistently, report the most.

Boardroom awareness lags in a similar pattern. Only 55% of boards are reported to have a clear, accurate view of the risk and security threats facing their organization, a figure that drops to 44% at smaller companies with under $5 billion in revenue. Decision-making about which risk technologies get funded tells a related story: CTOs, CEOs, and CIOs hold outsized influence over risk technology purchases, 80%, 76%, and 72% respectively report high influence, while CISOs sit meaningfully lower at 58%, and Chief AI Officers lower still at 40%, a gap that puts the people closest to the risk furthest from the budget conversation.

Investment priorities show a similar mismatch. Cloud security and operational risk tools are seeing the strongest planned spending increases, both above 13% average growth. But AI-driven threat detection, arguably the fastest-growing risk category by every measure in the same report, is seeing the lowest average investment increase of any area tracked, at just under 11%. Organizations appear to recognize AI as a rising threat faster than they're funding the tools built to catch it.

Governance is where most organizations stall

Gartner has predicted that more than 40% of agentic AI projects will be canceled by the end of 2027, driven by escalating costs, unclear business value, or inadequate risk controls, not by the technology failing to work. In many organizations, AI security failures won't trace back to a technical gap; they'll trace back to nobody having decided who's in charge, and of what, until it was too late.

Before deploying AI agents at scale, ServiceNow frames the governance question around what it calls the "ACID" test:

Autonomy. Set clear boundaries for which actions an agent can take independently, which require human approval, and which stay off-limits entirely. Elevated access and privilege escalation capabilities should never be handed to an agent by default; those are exactly the capabilities attackers target first, because compromising one unlocks broad system access.

Classification. Not every agent request carries equal weight. Governance frameworks should route routine, low-risk actions automatically while escalating higher-risk requests to a human decision-maker, and that logic needs to be designed into the architecture from the start rather than added after an incident forces the question.

Identity. Every agent needs a verifiable identity, credentialed and monitored the same way a new hire would be, not handed broad access on day one. This is the direct link back to why Veza's Access Graph matters: an organization can't govern an agent's identity if it can't first see what that identity is permitted to do.

Detection. Behavioral baselines need to exist so that anomalies, an unexpected spike in API calls, an unusual data access pattern, a sudden change in prompt length, trigger an alert rather than going unnoticed. Security shifts from defending a perimeter to continuously monitoring for drift inside the enterprise.

Getting this right also isn't a job for a single executive or an advisory committee that produces recommendations nobody's bound to follow. Organizations that see measurable AI benefits tend to be the ones that established governance councils with genuine, binding executive authority, spanning business units, C-suite leadership, and technical teams, rather than a committee that meets quarterly and files a report nobody reads.

The practical questions worth asking before you scale

A handful of questions tend to separate organizations that scale AI-driven security programs successfully from those that stall:

How are you monitoring AI activity today? Most organizations discover they're tracking what's easy to measure, not what matters most. Traditional tools watch network traffic and user logins; agents behave differently, and the blind spots between existing dashboards are usually larger than assumed.

Which workflows genuinely justify the investment agents require? High-volume, high-exposure workflows like fraud detection or customer triage at scale justify full identity verification and continuous monitoring. Lower-visibility workflows can become blind spots for exactly the kind of shadow AI usage that's already responsible for a meaningful share of AI-related breaches.

Who is accountable when an agent gets something wrong? If a security team gets blamed for a decision a business unit deployed without their sign-off, that's a structural problem, not a technology one. Accountability needs to be mapped before deployment, not reconstructed after an incident.

Where does governance authority sit, and what restructuring does that require? Centralized governance is more consistent but can slow deployment; distributed governance moves faster but risks gaps. There's no universally right answer, but the tradeoff needs to be a deliberate choice rather than something that happens by default.

How much deployment speed are you willing to trade for security, and at what cost? Conservative, heavily gated rollouts protect against incidents but can cost market position to faster-moving competitors. Aggressive rollouts move fast but accept higher incident probability. Making that tradeoff explicit beats letting it drift.

Getting the rollout right

Organizations that get real value from this part of ServiceNow tend to follow a similar sequence, and it isn't a full platform overhaul on day one. In the first six months, the priority is establishing visibility and control: build a complete inventory of AI agents and register them, stand up a governance council with real authority within 30 days, define which actions agents can take independently versus which require approval, and remove hardcoded credentials and API keys from any data an agent touches. From six to 12 months, the focus shifts to monitoring and identity infrastructure: launch a centralized control tower view, deploy behavioral monitoring across API calls and execution patterns, and enforce least privilege rigorously so agents only ever hold the access their specific task requires. From year two onward, organizations layer in more advanced capabilities: guardian agents that monitor other AI systems, risk-based validation where scrutiny scales with stakes, and alignment with emerging standards like ISO 42001 for AI system lifecycle management.

That sequencing matters more than any single tool choice. A Configuration Management Database that's genuinely trusted, a control library that's kept current, and a governance council with real teeth tend to determine whether a rollout compounds or stalls, more than which specific product gets deployed first.

"Our ServiceNow project was not an implementation, it was a transformation. Risk became visible and action became possible."

Andreea Obreja, Platform Owner, Vienna Insurance Group

Where to start

The gap between what most people picture when they hear "ServiceNow security" and what's actually there today is wider than it's ever been, and it's widening fast given how much has shipped in the past eight months alone. Buying a single piece, a vulnerability scanner integration here, a compliance module there, gets you a narrower slice of value than the platform is built to deliver. The compounding value shows up when security operations, risk and compliance, identity governance, and asset visibility all draw from the same CMDB and the same AI Control Tower, so a vulnerability, a vendor risk, and an AI agent's permissions become three views into one system rather than three separate problems three teams are solving in isolation.

EnableNow helps organizations figure out where they stand today before recommending any specific product, whether that means strengthening governance and audit workflows through ServiceNow Integrated Risk Management or unifying threat detection and response through ServiceNow Security Operations Management. If you're trying to work out what this should look like for your organization, reach out to EnableNow for an assessment.

FAQs

Is ServiceNow IRM the same thing as ServiceNow GRC? Largely yes. Integrated Risk Management is the current name for the product line long known as GRC, reflecting a shift from document-based risk tracking toward controls wired directly into live platform data, like the CMDB and incident records, rather than a rename of an unrelated product. ServiceNow's own materials still refer to the pairing as "IRM/GRC."

Does adopting ServiceNow here mean replacing our SIEM or existing security tools? No. Security Incident Response, Unified Security Exposure Management, and Threat Intelligence Security Center are all built to ingest data from existing SIEMs, EDR platforms, and scanners rather than replace them, adding CMDB context and orchestration on top of tools already in place.

Is Third-Party Risk Management included automatically with Integrated Risk Management? No. ServiceNow licenses Third-Party Risk Management separately from the core IRM packages, which surprises some buyers who assume it's bundled in from the start. It's worth confirming scope explicitly during procurement.

How long does a typical implementation take? A first release scoped to a single framework or a specific business unit commonly lands in 10 to 16 weeks. The two biggest variables affecting that timeline are the condition of the existing CMDB and how much of the control library already exists in a usable form before the project starts.