
A new report has revealed that the number of vulnerabilities within artificial intelligence services surged tenfold in the first half of 2026, highlighting the security risks that accompany the rapid adoption of AI technologies. The report, compiled through a coordinated disclosure programme and shared with relevant technology providers, documented a total of 188 vulnerabilities affecting 22 AI products from 20 vendors . This represents a sharp increase from the previous two years, with the number of reported vulnerabilities roughly doubling in 2025 compared to 2024 before exploding in 2026 . The findings reflect the growing reliance on AI systems across industries, as well as the increased scrutiny these systems are now facing from both security researchers and malicious actors .
Key features of the report include the finding that the median severity of the detected vulnerabilities was medium, but the pace at which they are being identified is a significant concern . A total of 152 vulnerabilities were publicly reported and documented during the first half of 2026, with the remaining cases either awaiting publication or containing highly sensitive information . The report emphasized that many vulnerabilities in AI services are not entirely new but are built into the foundational frameworks, libraries, and dependencies that developers rely on . This means that insecure code from upstream components can cascade down the stack, exposing AI products built on top of vulnerable frameworks . The rising number of security issues in AI systems aligns with warnings from the GSM Association about Africa’s readiness for an AI-driven future, highlighting the urgent need for robust cybersecurity frameworks .
The report comes amid heightened global attention on AI safety and governance . In June 2026, the United Kingdom announced a £30 million investment to establish an AI Safety Institute in San Francisco, reflecting growing concerns about the risks of frontier AI . AI-enabled cyberattacks are evolving rapidly, and concerns are mounting that the speed of AI adoption is outpacing the development of safety protocols . The report highlighted that many vulnerabilities in AI services are straightforward in nature and often result from inadequate oversight of third-party software and underlying dependencies . The rapid increase in reported vulnerabilities underscores the need for organisations to integrate security into the entire lifecycle of AI development and deployment, rather than treating it as an afterthought .

