Google GTIG: AI-Discovered Flaws Lead to RCE Twice as Often, and Agent Frameworks Are the Largest Layer of AI-Stack CVEs

The software ecosystem at large, and the AI stack in particular

Google Threat Intelligence Group published "Vulnerability Discovery and Exploitation Trends in the AI Era," covering January 2025 to August 2026. Monthly CVE disclosures doubled from 5,045 in January 2026 to 10,740 in August. 50% of likely AI-discovered vulnerabilities lead to remote code execution, against 26% of other CVEs. Exploited vulnerabilities reached 141 in January-August 2026, against 127 in all of 2025, and zero-days rose from 8 to 11 per month. GTIG counted 2,076 vulnerabilities in AI-related software, more than 1,500 of them disclosed in 2026, with agent orchestration frameworks such as Flowise, Langflow, and LangChain the largest layer at 782. BeyondTrust CVE-2026-1731, found autonomously by an AI research agent, was exploited within four days of disclosure and by five threat clusters within seven

Discovery has been automated and weaponization follows within days, while remediation still runs on human schedules. The agent orchestration layer, which holds credentials and executes code, is where AI-stack vulnerabilities concentrate