AI Exposure Surges as Vulnerabilities Jump 51%

The rapid adoption of generative AI, open-source components and automated development workflows is widening the corporate attack surface beyond traditional networks. Cybersecurity firm Forescout said threats are increasingly reaching model services, software dependencies and delivery infrastructure, making accurate asset inventories and continuous monitoring more critical as companies deploy AI systems and rely on interconnected development tools.

More than 37,000 vulnerabilities were disclosed globally in the first half of 2026, a 51% increase from a year earlier, according to Forescout’s latest report. Software supply-chain attacks expanded beyond package repositories into CI/CD pipelines and tools including LiteLLM. Meanwhile, the number of internet-exposed AI services surged 124% to 940,000, increasing the pool of systems that attackers can scan, compromise and exploit.