Your Organization's AI Trust Infrastructure Is Failing, New Survey Warns

Enterprise leaders are dangerously overconfident about AI security even as autonomous agents breach data boundaries daily — and a sweeping new report reveals the gap between policy and reality is widening fast.

A landmark survey of 750 enterprise leaders across the Americas, EMEA, and APAC has found that organizations are rapidly losing control of their AI ecosystems, with nearly 9 in 10 companies reporting at least one AI agent-related security incident in the past year. The research, published June 30, 2026, by data governance firm AvePoint, exposes a systemic failure in how businesses are managing the explosive growth of autonomous AI agents embedded in their daily operations.

The findings arrive at a critical inflection point. AI agents — systems capable of executing multi-step workflows, calling APIs, and making autonomous decisions — have moved far beyond the experimental stage. Nearly half of enterprise employees (46.9%) now rely on them daily or weekly. Yet the governance frameworks meant to oversee these tools remain dangerously underdeveloped, leaving organizations exposed to data breaches, unauthorized access, and compounding regulatory risk. For organizations still navigating the structural and cultural barriers that slow responsible AI adoption, this governance deficit represents an urgent and growing liability.

---

The Visibility Crisis Enterprises Can No Longer Ignore

The most striking data point in AvePoint's report is the near-tripling of organizations that cannot detect whether employees are using unsanctioned AI tools. That figure jumped from 6.3% to 17.6% in a single year for general AI tools. For AI agents specifically, the blind spot climbs even higher — past 21%.

This rapid decay in visibility has real consequences. Traditional shadow IT discovery tools were built for a world of human users navigating static software environments. Autonomous agents operate differently — they call APIs, scrape internal databases, synthesize documents, and act on behalf of users, often without triggering conventional monitoring alerts.

"AI is now integrated into everyday operations across regions and sectors, but our report makes it clear that accelerating adoption is outpacing readiness. Nearly half of employees now rely on AI agents weekly or daily, but visibility into unsanctioned tools is weakening, and AI-related incidents remain widespread."
— Dana Simberkoff, Chief Risk, Privacy and Information Security Officer, AvePoint

The result is a governance surface that is expanding faster than security teams can map it. The study also found that 35.5% of all enterprise data is already AI-generated, with that figure projected to climb to 42.1% within the next 12 months. Organizations are now tasked with securing pipelines where data is created by AI, processed by autonomous agents, and stored in corporate repositories — often without any human validating access controls along the way.

Why Conventional Monitoring Tools Are No Longer Enough

Standard endpoint detection and network monitoring platforms were architected around a relatively predictable threat model: a human user interacting with a defined application. Autonomous agents fundamentally break that model. A single agent can authenticate across multiple services, generate and store documents, query internal knowledge bases, and relay outputs to third-party platforms — all within a single workflow execution, and none of it necessarily visible through a conventional security information and event management (SIEM) dashboard.

The implication is stark: organizations that have not yet modernized their detection infrastructure are operating with a structural blind spot that grows larger with every new agent deployed. Understanding the broader risks and governance challenges AI introduces into business operations is now a foundational requirement, not an optional consideration for security teams.

---

The Confidence Paradox Undermining Enterprise Security

Perhaps the most alarming finding in the report is what AvePoint terms the "confidence paradox." More than 4 in 5 organizations stated they are confident in their ability to prevent unauthorized AI-related data access. Yet up to 72% of that same confident group experienced an unauthorized data access incident in the past 12 months.

This disconnect stems from a legacy mindset: measuring security readiness by whether a policy exists rather than whether technical controls are operational, enforceable, and auditable. Publishing an AI acceptable-use policy is not the same as enforcing one.

"Adversaries are using AI to operate at a scale and speed that makes traditional, static defenses obsolete. The window between a vulnerability's discovery and its exploitation has shrunk from months to days — and soon it will be merely minutes."
— Chandra Gnanasambandam, CTO, SailPoint

Chris Radkowski, GRC Expert at Pathlock, added important context about scale. Machine identities — service accounts, AI agents, and automated workflows — now outnumber human users across the enterprise by 20 times. "MFA and legacy access controls were built for a world of human users, not autonomous agents," he said. "As agentic AI takes on real business actions with real permissions, the attack surface expands in ways most organizations aren't prepared to see, let alone secure."

The Hidden Cost of Delayed Deployment

The pressure is compounding from another direction. Nearly 9 in 10 companies report delaying AI deployments by an average of almost six months specifically because of data security and governance concerns. Security teams are navigating territory that changes faster than their instruments can track — and the organizational cost of that hesitation is measurable in lost competitive advantage, slower innovation cycles, and widening capability gaps relative to less cautious competitors.

Delay is not a neutral position. Every month an organization defers AI deployment while its governance infrastructure remains immature is a month spent accumulating technical debt that will be harder and more expensive to resolve later. The financial and operational dimensions of this challenge are explored further in the context of how AI is fundamentally reshaping enterprise operations and competitive strategy.

---

Building a Trust Infrastructure That Actually Works

Security experts interviewed for the report converged on a clear prescription: organizations must abandon static, policy-first thinking in favour of real-time, operationally enforced governance.

Nathaniel Jones, Vice President of Security and AI Strategy at Darktrace, framed the strategic stakes directly: "The organizations likely to perform best over time will be those that become better at prioritization, behavioral detection, attack-path analysis, and identifying operational anomalies earlier in the intrusion lifecycle — particularly before public indicators or broad industry awareness emerge."

Elad Luz, Head of Research at Oasis Security, focused specifically on non-human identities (NHIs). "If AI agents are assigned persistent, unmanaged service accounts, these identities can quickly become overprivileged and unmonitored, increasing the organization's attack surface," he warned. His recommendation: automated monitoring, enforced least privilege, and clear governance policies for AI-driven NHIs established early rather than retrofitted later. The NIST AI Risk Management Framework offers a structured starting point for organizations looking to formalise these controls across the AI lifecycle.

Three Operational Guardrails Security Teams Should Prioritise Now

AvePoint's report identified three operational guardrails that security teams should address immediately:

1. Continuous Automated Discovery
Move past static endpoint monitoring to intercept and catalog API calls tied to AI backends. Unsanctioned agents that never touch a managed device will remain invisible to conventional tools without this capability in place.

2. Dynamic, Data-Centric Permissions
Clean up internal data sharing before indexing content into enterprise AI search engines. Permissions assigned to human users will be inherited — and often amplified — by agents operating on their behalf.

3. Behavioral Guardrails
Monitor agent activity for anomalies, such as an unauthorized agent suddenly requesting large batches of sensitive HR or financial records. Behavioral baselining is the difference between detecting a breach in minutes and discovering it months later during an audit.

What Executive Leadership Must Do Differently

"Trust cannot depend on policy, optimism, or model capability alone. Organizations need enforceable governance, lifecycle controls, proactive data protection, and continuous visibility into the data AI can access, create, and act on."
— Dana Simberkoff, AvePoint

The report's findings carry direct, actionable implications across the leadership tier. Security and IT leaders should immediately audit which AI agents currently operate within their environments and what permissions those agents inherit. Procurement and governance teams should evaluate purpose-built AI Agent Management Platforms as a priority investment rather than a deferred consideration. Executive leadership should reframe AI deployment timelines around trust infrastructure readiness — recognizing that the six-month delays companies are already experiencing are a symptom of governance gaps that will only grow more costly the longer they go unaddressed.

The organizations that will lead in the agentic AI era are not necessarily those that move fastest — they are those that move with the clearest visibility into what their AI systems are doing, and with the infrastructure in place to act when something goes wrong.