Resilience Finds AI Amplifies Familiar Cyberattacks in H1 2026 Claims

Resilience's 2026 Midyear Cyber Risk Report found that 85.3% of incurred losses in its insurance portfolio during H1 2026 traced to social engineering, up from 17.7% in H1 2024, while recording no incurred losses from AI-specific attack vectors. The report indicates AI's primary financial impact is amplifying familiar deception attacks rather than creating a new class of insured loss.

Resilience Finds AI Amplifies Familiar Cyberattacks in H1 2026 Claims Resilience's 2026 Midyear Cyber Risk Report says 85.3% of incurred losses in its insurance portfolio during the first half of 2026 traced to social engineering, up from 17.7% two years earlier. The company recorded no incurred losses from AI-specific attack vectors, suggesting that AI's clearest financial effect in its portfolio is to strengthen familiar deception rather than create a separate class of insured loss. Resilience's 2026 Midyear Cyber Risk Report says AI is showing up in its insurance claims by making familiar attacks more convincing and scalable, not by creating a separate wave of AI-native insured losses. The company analyzed claims from its own portfolio alongside intelligence from its Risk Operations Center; its findings describe that book of business, not all cyber incidents worldwide. Human deception dominates the loss mix In the first half of 2026, Resilience attributed 85.3% of incurred losses to social engineering as the point of failure, up from 17.7% in the first half of 2024. The report describes victims believing a false voice, message or request, and says attackers are using AI to sharpen those established techniques. The same report says social engineering represented 21.7% of claim frequency in H1 2026. That gap matters: the 85.3% figure is a share of incurred loss value, not a claim count or an estimate of the global breach market. Resilience recorded no incurred losses in the period from attack vectors it classified as AI-specific, including prompt injection, model exploitation or agentic-AI misuse. That does not mean those attacks do not exist. It means the company's H1 claims data had not yet identified them as a distinct driver of insured loss. Ransomware remains rare but costly Extortion, driven by ransomware, accounted for 73.05% of incurred losses while representing 5.8% of claims. Resilience also reports that immutable-backup adoption among clients that self-reported controls rose from 79.9% to 85.2% year over year. The company says that improvement is likely one reason claims remain relatively infrequent, but the report does not establish a causal effect. Vendor-related incidents accounted for 2.3% of incurred losses in H1 2026, down from 33.5% a year earlier. Resilience cautions that the lower share reflects the absence of a broad, high-severity vendor event in the period, not the disappearance of third-party exposure. What security teams can act on The report recommends phishing exercises that reflect AI-enabled deception, callback and dual-approval checks for sensitive transactions, continuous monitoring for compromised credentials, and contingency planning for critical vendors. For practitioners, the useful signal is operational: test whether controls contain a successful deception attempt and limit financial loss, rather than treating AI as a wholly separate attack category. Key Points - 1Resilience attributes 85.3% of H1 2026 incurred losses in its portfolio to social engineering as the point of failure, compared with 17.7% in H1 2024. - 2The company recorded no incurred losses from AI-specific attack vectors in the period; that portfolio finding is not evidence that such attacks are absent globally. - 3Extortion represented 73.05% of incurred losses but only 5.8% of claims, while immutable-backup adoption rose to 85.2% among clients that self-reported controls. Scoring Rationale The report provides current claims-based evidence about how AI changes attacker economics and insured loss severity, with direct relevance to security and risk teams; its scope is limited to Resilience's portfolio and self-reported controls, so the finding is material but not industry-wide proof. Sources Primary source and supporting public references used for this report. Practice with real Health & Insurance data 90 SQL & Python problems · 15 industry datasets 250 free problems · No credit card See all Health & Insurance problems /problems/datasets/health