AI-Driven Software Glitches Spark Outages but Exploited Vulnerabilities Remain Rare

In July 2026 a series of AI‑related software failures disrupted services worldwide. Payment card networks in Japan stopped, an ignored security patch knocked out internet access for millions in Australia, United Airlines halted flights across the United States, an Azure cloud outage affected western U.S. regions, and a hospital information system in New Zealand failed, among other incidents. These events illustrate that AI‑augmented systems and outdated updates can still cause significant operational damage.

A separate VulnCheck analysis of 1,061 AI‑identified security flaws found only 14 (about 1.3 %) were used in real attacks, a rate virtually identical to the overall exploitation rate for known vulnerabilities. While AI now helps researchers discover many more potential flaws, attackers have not yet turned those discoveries into a surge of successful cyber‑attacks, keeping AI’s impact on real‑world exploitation limited.