One in Four Data Breaches Is Now AI-Enabled, and 2026 Is Already a Record Year

Between March 2025 and February 2026, one in four data breaches was AI-enabled — up 56% from a year earlier. The Identity Theft Resource Center tracked 1,803 data compromises in the first half of 2026 alone, putting the year on pace to surpass 2025's record of 3,321 incidents, with more than 471 million victim notices already issued in H1 — above all of 2025's 297.5 million.

One Breach, 58% of the Total

A single breach involving Instructure Holdings' Canvas education platform accounted for an estimated 275 million of H1 2026's victim notices — roughly 58% of the total. Supply-chain attacks separately generated another 280.6 million victim notices from just 38 initial breach events, underscoring how concentrated modern breach impact has become: a small number of upstream compromises now cascade into an outsized share of total exposure.

What 'AI-Enabled' Actually Means Here

This is not primarily about AI models being hacked — it's attackers using AI to scale reconnaissance, generate more convincing phishing and social-engineering content, and automate parts of the intrusion chain that previously required more manual attacker effort. The result is more attacks launched per attacker, not necessarily smarter individual attacks.

• A single well-executed supply-chain or platform-level breach can now generate victim notices in the hundreds of millions — vendor risk assessment matters more than ever
• AI-enabled reconnaissance and phishing generation means employee security training needs to assume more convincing, more personalised attack attempts than a year ago
• Breach volume growth at this rate is itself a budget argument — security tooling and incident-response readiness should scale with genuine attacker capability, not last year's threat model