MITRE ATLAS Attack Landscape

4,226 AI-related CVEs analyzed → 12,234 mappings across 101 ATLAS techniques (each CVE may match multiple techniques).

4,226
AI CVEs
101
Techniques
12,234
Total Mappings
Exploit Public-Facing Application
#1 (2615 CVEs)
CISO AnalysisData updated 2026-11-18

Executive Summary

The AI attack landscape is dominated by a small set of high-volume techniques. Across 4,410 AI-related CVEs mapped to 170 MITRE ATLAS techniques, Exploit Public-Facing Application (AML.T0049) leads with 2,615 mapped CVEs — reflecting the reality that most AI/ML systems are deployed behind web APIs with insufficient input validation. AI Agent Tool Invocation (1,077), AI Software (1,029), Valid Accounts (813), and Denial of AI Service (792) round out the top five.

The concentration is meaningful: the top 5 techniques account for 39.4% of all CVE-to-technique mappings, while the long tail spans more than 165 techniques with much sparser coverage. Security teams can achieve disproportionate risk reduction by focusing detection and response on a small set of attack patterns — rather than spreading resources thin across the full ATLAS matrix.

Key Findings

  • Initial Access dominates the tactic ranking with 2,980 unique CVEs — public-facing exposure plus weaknesses in the software stack around models drive this category. Privilege Escalation (1,730) and Execution (1,628) follow.
  • AI Agent Tool Invocation is the second-largest technique with 1,077 CVEs (AML.T0053). This category is essentially the shadow attack surface around the model: deserialization in pickle files, RCE in inference servers, unsafe deserializers in agent frameworks.
  • AI Software is more prevalent than commonly assumed. 1,029 CVEs target this surface. Most AI incident response plans don't cover availability attacks at all.
  • 25 AI CVEs are in CISA's KEV catalog — actively exploited in the wild. They span inference servers (Ollama, vLLM), MLOps platforms (MLflow), and UI frameworks (Gradio).
  • 1,777 AI CVEs (40%) have public exploit code available — almost half of the AI CVE landscape has weaponized PoCs, dramatically shortening the window between disclosure and active exploitation.
  • Growth is steady. 408 new AI-related CVEs were added in the last 30 days alone, confirming the threat surface is expanding faster than most security programs adapt.

Trend Analysis

The shift from model-level attacks (adversarial examples, jailbreaks) toward infrastructure-level exploitation marks a maturation of the AI threat landscape. The data is unambiguous: the real attack surface is the software stack around the models — frameworks, APIs, serving infrastructure, data pipelines.

Agent frameworks remain the emerging frontier. As AI systems gain tool-use capabilities (file access, code execution, web browsing), each tool integration becomes a potential attack vector. Agent-related CVEs continue to grow in both volume and severity, with many enabling remote code execution through prompt injection chains that pivot into the underlying tool runtime.

The patching picture is more nuanced than the early "crisis" narrative suggested. Across all AI package CVE associations, 40.0% have a documented fix available — better than initially feared, but still well below the 60-70% rate typical of the broader software ecosystem. The gap between AI tooling and mainstream software security maturity is real but narrowing.

Recommendations

  1. Prioritize the top 5 ATLAS techniques for detection engineering. Build detection rules specifically for the leading techniques shown above. Together they cover 39.4% of the threat landscape.
  2. Audit your AI supply chain. Inventory all AI/ML dependencies, check against our package risk scores, and establish a vetting process for new framework adoption. Pay special attention to packages with risk scores above 70 (PyTorch, Ollama, MLflow, Gradio, LiteLLM, LangChain, LangFlow).
  3. Implement input validation at every AI system boundary. The dominance of "Exploit Public-Facing Application" and "AI Agent Tool Invocation" mappings means robust input sanitization at API endpoints, model inputs, and agent tool interfaces delivers the highest security ROI.
  4. Monitor CISA KEV for AI-specific entries. The 25 AI CVEs currently in KEV should be patched within CISA's remediation timelines. Set up automated alerts for new AI KEV additions.
  5. Plan for AI system availability attacks. Include resource exhaustion, recursive loops, and inference overload in incident response playbooks. Most organizations lack AI-specific DoS detection.

Methodology

This analysis is based on 4,410 AI-related CVEs tracked by AI Threat Alert, mapped to 170 MITRE ATLAS techniques via automated enrichment (Claude AI) and manual validation. Technique frequency reflects the number of distinct CVEs mapped to each technique — a single CVE may map to multiple techniques. Tactic counts reflect distinct CVEs mapped to any technique under that tactic. Data sources include NVD, GitHub Security Advisories, CISA KEV, EPSS, OSV, and vendor advisories. All numeric values in this analysis are pulled live from the database on every page load — they always match the charts and tables below.

#TechniqueCVEs
1AML.T0049Exploit Public-Facing Application2615
2AML.T0053AI Agent Tool Invocation1077
3AML.T0012Valid Accounts813
4AML.T0029Denial of AI Service792
5AML.T0050Command and Scripting Interpreter699
6AML.T0055Unsecured Credentials570
7AML.T0025Exfiltration via Cyber Means492
8AML.T0037Data from Local System472
9AML.T0107Exploitation for Defense Evasion343
10AML.T0083Credentials from AI Agent Configuration339
11AML.T0081Modify AI Agent Configuration327
12AML.T0040AI Model Inference API Access267
13AML.T0011User Execution245
14AML.T0034Cost Harvesting233
15AML.T0086Exfiltration via AI Agent Tool Invocation233
16AML.T0072Reverse Shell183
17AML.T0106Exploitation for Credential Access177
18AML.T0036Data from Information Repositories157
19AML.T0105Escape to Host156
20AML.T0058Publish Poisoned Models150
21AML.T0006Active Scanning142
22AML.T0035AI Artifact Collection133
23AML.T0074Masquerading116
24AML.T0075Cloud Service Discovery116
25AML.T0084Discover AI Agent Configuration103
26AML.T0078Drive-by Compromise96
27AML.T0085Data from AI Services88
28AML.T0098AI Agent Tool Credential Harvesting80
29AML.T0080AI Agent Context Poisoning79
30AML.T0101Data Destruction via AI Agent Tool Invocation78
31AML.T0043Craft Adversarial Data67
32AML.T0112Machine Compromise52
33AML.T0057LLM Data Leakage44
34AML.T0007Discover AI Artifacts43
35AML.T0051LLM Prompt Injection43
36AML.T0110AI Agent Tool Poisoning34
37AML.T0070RAG Poisoning32
38AML.T0108AI Agent32
39AML.T0020Poison Training Data31
40AML.T0052Phishing31
41AML.T0093Prompt Infiltration via Public-Facing Application29
42AML.T0097Virtualization/Sandbox Evasion29
43AML.T0079Stage Capabilities28
44AML.T0091Use Alternate Authentication Material28
45AML.T0021Establish Accounts26
46AML.T0031Erode AI Model Integrity26
47AML.T0073Impersonation25
48AML.T0059Erode Dataset Integrity24
49AML.T0064Gather RAG-Indexed Targets23
50AML.T0018Manipulate AI Model22
51AML.T0099AI Agent Tool Data Poisoning21
52AML.T0102Generate Malicious Commands21
53AML.T0010AI Supply Chain Compromise20
54AML.T0096AI Service API20
55AML.T0087Gather Victim Identity Information19
56AML.T0104Publish Poisoned AI Agent Tool18
57AML.T0024Exfiltration via AI Inference API17
58AML.T0047AI-Enabled Product or Service15
59AML.T0056Extract LLM System Prompt15
60AML.T0092Manipulate User LLM Chat History14
61AML.T0001Search Open AI Vulnerability Analysis11
62AML.T0076Corrupt AI Model11
63AML.T0044Full AI Model Access9
64AML.T0100AI Agent Clickbait9
65AML.T0019Publish Poisoned Datasets8
66AML.T0109AI Supply Chain Rug Pull8
67AML.T0054LLM Jailbreak7
68AML.T0077LLM Response Rendering7
69AML.T0041Physical Environment Access6
70AML.T0065LLM Prompt Crafting6
71AML.T0014Discover AI Model Family5
72AML.T0015Evade AI Model5
73AML.T0063Discover AI Model Outputs5
74AML.T0069Discover LLM System Information5
75AML.T0066Retrieval Content Crafting4
76AML.T0068LLM Prompt Obfuscation3
77AML.T0103Deploy AI Agent3
78AML.T0002Acquire Public AI Artifacts2
79AML.T0071False RAG Entry Injection2
80AML.T0082RAG Credential Harvesting2
81AML.T0094Delay Execution of LLM Instructions2
82AML.T0111AI Supply Chain Reputation Inflation2
83AML.T0000Search Open Technical Databases1
84AML.T0003Search Victim-Owned Websites1
85AML.T0013Discover AI Model Ontology1
86AML.T0042Verify Attack1
87AML.T0046Spamming AI System with Chaff Data1
88AML.T0048External Harms1
89AML.T0061LLM Prompt Self-Replication1
90AML.T0089Process Discovery1
91AML.T0090OS Credential Dumping1
92AML.T0095Search Open Websites/Domains1
93AML.T0004Search Application Repositories0
94AML.T0005Create Proxy AI Model0
95AML.T0008Acquire Infrastructure0
96AML.T0016Obtain Capabilities0
97AML.T0017Develop Capabilities0
98AML.T0060Publish Hallucinated Entities0
99AML.T0062Discover LLM Hallucinations0
100AML.T0067LLM Trusted Output Components Manipulation0
101AML.T0088Generate Deepfakes0

Track these techniques against your AI stack with real-time alerts.

Start 14-Day Free Trial