Report by Google Threat Intelligence Group

Vulnerability Discovery and Exploitation Trends in the AI Era

35 FINDINGSPublished Oct 1, 2026
View Original Report →

Key Findings

Average monthly vulnerabilities exploited increased from 10.5 per month in 2025 to 18 per month from January 2026 to August 2026.

ExploitationVulnerabilities

Average monthly zero-day exploitations increased from 8 per month in 2025 to 11 per month from January 2026 to August 2026, peaking at 22 in August 2026.

Zero-DayExploitation

Baseline High-Risk disclosures more than doubled from about 65 per month in mid-2025 to about 135 per month in mid-2026.

VulnerabilitiesThreat Trends

Mass research disclosures against TOTOLINK consumer router firmware added 75 High-Risk flaws in April–May 2026.

VulnerabilitiesNetworking

Only 0.23% of all disclosed vulnerabilities in 2026 (about 1 in 431) were observed in active exploitation.

ExploitationVulnerabilities

Overall CVE disclosure growth indexed +128% compared to baseline (January 2025).

VulnerabilitiesTrend Analysis

High-Risk vulnerabilities disclosed increased by 241% (about 3.5× baseline) by August 2026.

High RiskVulnerabilities

GTIG tracked 2,076 cumulative AI-related CVE disclosures from January 2025 through August 2026, with over 1,500 identified from January 2026 to August 2026 alone.

AI SecurityVulnerabilitiesAI Vulnerabilities

Orchestration and agent frameworks accounted for 782 AI-related vulnerabilities in 2026.

AI SecurityOrchestrationAI Vulnerabilities

AI web apps and portals accounted for 230 AI-related vulnerabilities in 2026.

AI SecurityWeb ApplicationsAI Vulnerabilities

Inference and serving infrastructure accounted for 212 AI-related vulnerabilities in 2026.

AI SecurityInferenceAI Vulnerabilities

Model security advisories accounted for 106 AI-related vulnerabilities in 2026.

Model SecurityAI SecurityAI Vulnerabilities

ML frameworks and hubs accounted for 99 AI-related vulnerabilities in 2026.

ML FrameworksAI SecurityAI Vulnerabilities

MLOps and experiment tracking systems accounted for 39 AI-related vulnerabilities in 2026.

MLOpsAI SecurityAI Vulnerabilities

Vector databases and search systems accounted for 19 AI-related vulnerabilities in 2026.

DatabasesAI SecurityAI Vulnerabilities

Backend serving infrastructure disclosures reached 212 vulnerabilities in 2026, with nearly 24% of these flaws stemming from unauthenticated API endpoints or SSRF.

InferenceAPI Security

AI-discovered vulnerabilities produced Information Disclosure at 8%, compared to 18% for vulnerabilities not identified as discovered by AI.

Information DisclosureAI Security

AI-discovered vulnerabilities produce Data Manipulation at 5%, compared to 9% for vulnerabilities not identified as discovered by AI.

Data ManipulationAI Security

Zero-day exploitations constituted 62% of all observed exploited vulnerabilities from January 2026 to August 2026.

Zero-DayExploitation

Monthly vulnerability disclosures doubled, rising from 5,045 in January 2026 to 10,740 in August 2026.

Vulnerabilities

High-Risk vulnerabilities represented 3% of all vulnerabilities disclosed in August 2026.

VulnerabilitiesRisk Assessment

Oracle and Linux advisories contributed 128 High-Risk vulnerabilities in August 2026, nearly 37% of all High-Risk disclosures that month.

VulnerabilitiesOracleLinux

From January 2026 to August 2026, 141 distinct vulnerabilities were disclosed and exploited, surpassing 127 exploited vulnerabilities for the full year 2025.

ExploitationVulnerabilities

Since May 2026, CVE exploitation showed +127% indexed growth closely mirroring disclosure growth of +128% indexed growth.

VulnerabilitiesThreat Trends

Exploitation of High-Risk vulnerabilities more than doubled from 28 in 2025 to 75 from January 2026 to August 2026.

ExploitationHigh Risk

Vulnerabilities affecting Edge and Security Appliances represented 14% of vulnerabilities exploited from January 2026 to August 2026, while 11% affected Enterprise Directory & Collaboration hubs.

EdgeEnterprise Security

Over 65% of edge flaws exploited from January 2026 to August 2026 met High/Critical Threat Risk ratings.

EdgeThreat Risk

Frontier models accounted for 97 AI-related vulnerabilities in 2026.

Frontier ModelsAI SecurityAI Vulnerabilities

AI-discovered vulnerabilities comprised 58% Medium-Risk, 39% Low-Risk, and 4% High-Risk vulnerabilities (Jan–Aug 2026).

AI SecurityRisk Assessment

Conventional (non-AI) CVE disclosures comprised 69% Low-Risk, 28% Medium-Risk, and 3% High-Risk vulnerabilities (Jan–Aug 2026).

VulnerabilitiesRisk Assessment

Orchestration middleware accounted for 50% of all AI-related flaws and experiences a +347% surge in disclosures in 2026.

Orchestration MiddlewareAI SecurityAI Vulnerabilities

Exactly 50% of AI-discovered vulnerabilities resulted in Remote Code Execution (RCE), compared to 26% across the broader CVE ecosystem.

Remote Code ExecutionAI Security

High-Risk vulnerability disclosures surged from 131 in January 2026 to 350 in August 2026, a 167% growth.

VulnerabilitiesRisk Management

Excluding Linux, Oracle, and Totolink, High-Risk disclosure growth was 128% from January 2025 to August 2026.

VulnerabilitiesVendor Risk

CVE exploitation in the wild increased by approximately 127% (indexed to January 2025).

ExploitationVulnerabilities