Google says attackers are using AI agents to automate differential patch analysis and the exploitation of n-day vulnerabilities, meaning publicly known security flaws. These workflows allow them to scale operations against internet-facing servers with little or no human intervention. The average number of exploited zero-day vulnerabilities rose from eight per month in 2025 to 11 per month from January through August 2026, before reaching 22 in August. Anthropic also reported multi-agent campaigns capable of automating much of the attack cycle, from vulnerability discovery to data theft.
Source: Read the original article


