Nearly 82% of executives surveyed by AXIS Capital Holdings Limited said their organizations plan to increase their cybersecurity budgets over the next 12 months, while 75.2% said they are likely to reduce cybersecurity headcount due to greater productivity resulting from investment in AI cybersecurity tools.

AXIS surveyed 500 chief executive officers (CEOs) and chief information security officers (CISOs) across the U.S. and UK, revealing striking differences in how executive leaders view the risks and rewards of AI.

“AI is clearly a transformative force for data analytics, innovation and operational efficiency, and it is also undeniable that AI is quickly propelling us toward an entirely new risk landscape,” said Vince Tizzio, AXIS president and CEO. “The findings in this report underscore our new environment, where a technology that promises untold productivity gains is also creating unprecedented risk.”

AXIS said that AI is improving cyber defense technologies, but it is also equipping cybercriminals with tools that can make threats faster, more adaptive and harder to detect.

In fact, executives across both regions identify AI-driven attacks (25.2%) as the top emerging cyber threat over the next 12 months, outranking identity theft (18.0%) and supply-chain compromise (16.6%).

Respondents also ranked their greatest AI-related risks:

• Data Leakage: Unauthorized exposure of sensitive information outside its intended environment (23.2%).
• Shadow AI: Unsanctioned use of AI tools by employees without IT/security safeguards or approval (21.6%).
• Model Manipulation: Deliberate tampering with an AI model to alter its behavior or outputs (19.0%).
• Deepfake/Social Engineering: Fake/AI-generated content used to deceive audiences and/or trick people into revealing information/taking harmful actions (17.6%).
• Regulatory Noncompliance: Failure to meet legal or industry rules governing data, security, or AI use (17.4%).

Overall, CEOs identified data leakage (28.7%) as their greatest AI-related threat, compared to 17.2% of CISOs. Meanwhile, CISOs ranked Shadow AI as their No. 1 risk (27.2%), highlighting fears over unmonitored and unregulated tools, compared to 16.5% of CEOs.

What keeps leaders up at night isn’t just the attack itself but the reputational damage (39.4%) and customer attrition (38.8%) that could follow.

The study also found that UK respondents are more cautious about AI adoption than their U.S. counterparts:

• Only 55.3% of UK CEOs said they were confident AI will strengthen their organization’s cyber defenses over the next three years compared to 88.4% of U.S. CEOs.
• U.S. leaders (82.8%) are also more likely to trust AI tools to help make cybersecurity decisions than those in the UK (43.3%). Overall, CEOs are more likely (67.1%) than CISOs (58.6%) to trust AI tools to help make cybersecurity decisions.
• In the U.S., 93.5% of CEOs and 87.5% of CISOs said they believed AI delivered return on investment for cybersecurity. In the UK, that belief came in at 69.1% among CEOs and 74% for CISOs.
• UK leaders feel less prepared (44%) for AI threats than those in the U.S. (84.8%).
• The survey also revealed that U.S. respondents (94%) are more likely to carry cyber insurance compared to UK respondents (68.4%).

“Our survey findings indicate the impact of AI in transforming corporate defense strategies while exposing differing views between CEO’s strategic optimism and CISO’s security prudence,” commented Lori Bailey, Head of Global Cyber and Technology at AXIS. “While it is now commonplace for CEOs to champion AI as a catalyst for innovation and efficiency, CISOs tend to see it as a new frontier of exposure and control.”

Methodology

Findings were derived from a 23-question online survey among a total of 500 CEOs and CISOs across the UK and U.S. In the U.S., the respondent pool consisted of 138 CEOs and 112 CISOs, whereas in the UK it consisted of 123 CEOs and 127 CISOs. Respondents represented companies with at least 250 employees. Fielding for this study was conducted by an independent company from Oct. 22-29, 2025.