The budget signal is unambiguous
PwC's latest Global Digital Trust Insights Survey, drawing on nearly 4,000 business and technology executives across 71 countries, finds 84% expect their organization's cyber budget to rise over the next 12 months, up from 78% in the prior survey cycle. That increase is not generic security anxiety, 58% of respondents specifically rank artificial intelligence among their top cyber budget priorities, which means the money is following a named cause rather than a vague sense that threats are worsening across the board.
A budget increase this broad and this clearly attributed to one cause, frontier AI rollout, should in theory make the accountability question simpler to answer inside any given organization. The survey's next finding shows that has not happened in practice, and the gap between rising spend and clear ownership is where this report earns its place on a CIO's reading list rather than a passing mention in a budget memo.
Three owners is the same as no owner
Asked who owns AI accountability, respondents split their answers almost evenly across three different roles: 29% point to the CIO or CTO, 26% to a dedicated AI leader such as a chief AI officer, and 17% to the CISO. No single function commands even a third of the vote, which in practice means most organizations represented in this survey do not have one clear owner for AI risk, they have three partial owners each covering a piece of the problem and likely assuming someone else is covering the rest.
That fragmentation is a structural problem, not a staffing gap that more headcount fixes on its own. A CIO focused on infrastructure and uptime, an AI leader focused on model performance and roadmap, and a CISO focused on threat detection and incident response are each optimizing for a different success metric, and none of those three metrics by itself captures the full risk surface an AI deployment actually creates. Budget increases poured into three partially overlapping, partially gapped ownership structures tend to fund redundant capability in some areas and leave other areas quietly uncovered, with no one positioned to notice the gap until something breaks.
Only 47% means the board is not actually watching
Just 47% of leaders strongly agree that cyber risk is a standing board-level agenda item, a figure that sits uncomfortably next to the 84% expecting budget increases and the three-way accountability split above it. Money is flowing up and ownership is fragmenting sideways, while board-level attention, the one mechanism that could force the fragmentation into a coherent structure, is present in barely half of represented organizations as a consistent, standing item rather than an occasional briefing.
This is the governance gap that tends to surface publicly only after an incident, when a post-mortem reveals that no one with sufficient authority had the full picture before something went wrong. Enterprises with budget rising, ownership split, and board attention inconsistent are assembling exactly the conditions that make a response slow and confused when an AI-related security incident actually occurs, and the root cause in that kind of post-mortem is almost always structural rather than personal, with every individual group having done its own piece correctly while no single group held the mandate or the complete view needed to act decisively across the whole picture.
Nobody wants to let the agent decide on its own
Only 22% of surveyed leaders would authorize fully autonomous AI agents to carry out cyber-defense actions on their own, and 50% specifically name AI-system attacks as the threat category they feel least prepared to handle. Read together, these numbers describe an industry racing to adopt agentic AI across the business while remaining conspicuously unwilling to trust the same category of technology to defend itself, a hesitation that is less a contradiction than it looks and more a reasonably calibrated read of current agent reliability under adversarial pressure.
That hesitation should inform how enterprise security teams pitch AI-powered defensive tooling internally this budget cycle. A vendor promising fully autonomous threat response is selling into a documented trust gap that most peer organizations have not yet closed, and pitching human-in-the-loop augmentation, where the agent proposes and a person approves, is likely to find a far more receptive board than a pitch built around removing the human step entirely, however much more efficient that framing sounds in a sales deck.
The basics still are not done either
Perhaps the most sobering detail in the whole survey is how far preparedness lags even on fundamentals that predate the AI conversation entirely. Only 39% of organizations have a fully formalized, integrated operational continuity plan specifically addressing cyber risk, and just 5% have fully implemented all seven data-risk measures PwC examined in the survey, with data classification and data-loss prevention each sitting around the halfway mark of full deployment. These are not cutting-edge AI governance challenges, they are foundational security hygiene that has been on every maturity checklist for a decade.
That gap matters because it reframes the AI risk conversation for any CISO or CIO building next year's roadmap. An organization that has not finished its basic data classification and continuity planning is not ready to layer sophisticated AI-specific governance controls on top of a foundation that is still under construction. The practical sequencing this survey argues for is unglamorous but necessary: close the fundamental data and continuity gaps first, assign a single clear owner for AI risk second, and only then invest in the more advanced AI-specific tooling that vendors are currently pitching hardest, because that tooling delivers far less value sitting on top of an incomplete foundation than it would once the basics are actually in place.
