IBM Study: One in Four Malicious Breaches are AI-Enabled, Costing Companies $6 Million on Average

IBM’s 2026 Cost of a Data Breach Report reveals that 25% of malicious breaches now involve AI tools, a 56% increase from the prior year. These breaches cost organizations an average of $6 million, nearly $1.01 million more than the global average of $4.99 million. The report attributes the rise to AI-powered deepfake scams and malware, which are becoming faster and cheaper to deploy. Companies using AI and automation in security operations reduced breach costs by nearly $2 million, yet 25% of organizations have not adopted these tools, widening the gap between attack and defense capabilities.

The study highlights a growing imbalance in cyber risk economics, where attacks can be launched for minimal cost while breach remediation expenses escalate. Organizations are shifting focus toward anticipating future threats rather than reacting to incidents. Ponemon Institute’s follow-up research found 85% of organizations plan to increase security spending after learning of advanced frontier AI cyber capabilities, compared to 64% after experiencing a breach. However, only 18% use AI agents for vulnerability management, leaving known weaknesses unaddressed despite shrinking exploit windows.

AI-driven attacks are disproportionately targeting critical infrastructure sectors, with 62% focused on industries like financial services and energy. Financial services breaches averaged $6.3 million, while energy breaches averaged $5.2 million. The concentration of attacks in these sectors raises concerns about systemic disruption to economies, supply chains, and essential services. The 2026 report, based on 602 breaches globally between March 2025 and February 2026, underscores the urgency for organizations to adapt security strategies to AI-driven threats.

IBM Security’s Suja Viswesan noted that AI is accelerating attack timelines while extending breach discovery and remediation gaps, directly inflating costs. The report recommends integrating remediation into development workflows, securing identities during runtime, and accelerating risk mitigation to match attacker speed. Conducted by Ponemon Institute and sponsored by IBM, the research reflects responses from 456 organizations in May 2026, with 78% aware of advanced frontier AI models such as Mythos.