The 2025-2026 SentinelOne AI and Cloud Verified Exploit Paths report identifies the adoption of AI and Large Language Models (LLMs) as the primary driver of modern cloud risk. With an observed 140% increase in AI-specific secrets, organizations are facing a rise in "shadow AI"—the unsanctioned use of unmanaged API keys. This sprawl enables unique attack vectors, including prompt injection, data poisoning, and unauthorized access to sensitive datasets processed by AI models.
Traditional security challenges also persist, with attackers frequently leveraging legacy vulnerabilities and misconfigured external services as initial entry points. The report emphasizes that high-privilege cloud provider keys and CI/CD tokens remain critical targets, potentially leading to large-scale data exfiltration and supply chain compromises. To counter these threats, security leaders are advised to implement continuous surface monitoring, automate DevSecOps workflows, and establish centralized governance for all AI-related credentials.
Top comments (0)