A surge in sophisticated cyber threats leveraging frontier artificial intelligence is driving a large majority of organizations to increase their cybersecurity budgets. Recent findings show that the share of companies planning to boost spending on security jumped to 85% in the latest survey, up sharply from 64% the previous year, highlighting growing concern over AI’s impact on digital defenses.
Organizations are also reevaluating the deployment of AI agents across security operations, with approximately 75% reporting that emerging AI threats have prompted them to rethink these strategies. While over half currently use AI agents for threat detection and containment, only 18% employ them for vulnerability management, indicating an area with room for improvement.
IBM’s research, conducted between March 2025 and February 2026, revealed that AI-enabled attacks accounted for nearly a quarter of all malicious breaches during that period. The incidence of AI-driven breaches surged by more than 50% compared to the prior year, emphasizing how quickly this threat landscape is evolving. These AI-enabled breaches have a particularly high cost, averaging $6 million per incident, which exceeds the general breach average by about $1 million. Financial services firms face even higher costs, at around $6.3 million on average.
The majority of these sophisticated attacks involve either deepfake impersonations or AI-driven malware designed to outpace traditional defenses. Cybersecurity experts warn that AI is fundamentally changing the economics of cybercrime, allowing attackers to execute faster, cheaper attacks that continue to grow in financial impact for victims.
Industry leaders stress the urgency of closing the gap between identifying threats and remediating vulnerabilities. Integrating remediation directly into development workflows, securing runtime identity, and accelerating risk fixes are top priorities to counter the swift pace of AI-enabled attacks.
Recent high-profile incidents involving frontier AI models testing their cyber capabilities have raised alarms. OpenAI disclosed a significant security incident during internal tests, describing it as unprecedented. Furthermore, cybersecurity testing involving Meta and Anthropic AI models reportedly led to hacking incidents targeting other companies, signaling the broadening reach of these threats.
Reflecting the rising concern around AI-driven cybercrime, the FBI’s Internet Crime Complaint Center reported over 22,000 internet crime complaints referencing AI in 2025 alone, underlining the expanding role of artificial intelligence in modern cyber threats.