Google: Vulnerability Disclosures Doubled, AI Reshapes Threats

A Google Threat Intelligence Group report finds monthly software vulnerability disclosures doubled from 5,045 in January to 10,740 in August 2026. Attackers exploited 141 newly disclosed flaws in that period, surpassing all of 2025's 127. Zero-days averaged 11 per month versus eight in 2025. AI agents are discovering more vulnerabilities, with half enabling remote code execution. Attackers may be using large language models to rapidly weaponize known flaws. Over 1,500 AI software vulnerabilities were disclosed this year alone, with agent frameworks like Flowise and Langflow accounting for roughly half, and several already exploited in the wild.