PwC Survey: Adversarial AI Tops Cyber Preparedness Gaps; Quantum and Agent-Autonomy Readiness Remain Low

PwC's 2027 Global Digital Trust Insights survey of 3,934 leaders in 71 countries names adversarial AI the biggest cybersecurity preparedness gap, with only 21% deploying quantum-resistant cryptography and only 22% willing to allow fully autonomous defense…

PwC's 2027 Global Digital Trust Insights report, published October 1, 2026, surveyed 3,934 business and technology leaders across 71 countries (SecurityWeek rounds this to nearly 4,000 leaders in more than 70 countries). Fifty-two percent identified adversarial AI attacks as their biggest cyber preparedness gap, with autonomous botnets (53%) and data poisoning (52%) also cited as threats organizations cannot adequately defend against; prompt injection remains an unsolved LLM risk. Accountability for AI risk is split across technology functions (29%), dedicated AI leaders (26%), and CISOs (17%). Data fundamentals lag: only 49% report full data classification and 48% full data-loss prevention. Readiness for AI autonomy and quantum threats is also low — only 22% would permit fully autonomous defensive agents, only 21% are deploying quantum-resistant cryptography against harvest-now-decrypt-later theft, and 44% of CISOs cite AI governance skills as a barrier to agent autonomy. Spending is expected to respond: 84% of security and finance leaders anticipate larger cyber budgets, and 58% rank AI highly among cyber spending priorities — though the two reports disagree on phrasing, with Infosecurity Magazine saying the 58% place AI among their top five priorities and SecurityWeek calling it the top priority.

• Survey covered 3,934 business and technology leaders in 71 countries (Infosecurity Magazine); SecurityWeek reports this as 'nearly 4,000' leaders in 'more than 70' countries.
• 52% of leaders call adversarial AI attacks their biggest cybersecurity preparedness gap.
• Threats leaders say they cannot adequately defend against: autonomous botnets (53%), adversarial attacks (52%), data poisoning (52%); prompt injection remains an unsolved LLM risk.
• AI risk ownership is split: technology functions 29%, dedicated AI leaders 26%, CISOs 17%.
• Only 49% report full data classification implemented; 48% report full data-loss prevention.
• Only 22% would permit fully autonomous defensive AI agents without human approval.
• Only 21% are deploying quantum-resistant cryptography against harvest-now-decrypt-later theft.
• 44% of CISOs cite AI governance skills as a barrier to granting agents autonomy.