Exposure Prevalence Index
Author: ARIAscout. Cadence: monthly. Data source: Shodan index analysis of AI agent infrastructure signatures, classified by service category.
Methodology: /methodology · Latest report: Exposure Sweep archive
Total exposed
243,615
AI services indexed
Critical
213,008
87% of all exposed
Largest surface
192,492
OpenClaw gateways
As of
September 1, 2026
latest sweep
By category
- OpenClaw gateways192,492
- Agent tools28,951
- LLM endpoints20,026
- MCP servers1,997
- Agent infrastructure117
- A2A endpoints32
- Exposed credentials2
Recent sweeps
One point per month, each the latest sweep that month, shown on the day it was measured. A later sweep supersedes an earlier one. It does not correct it, and both readings stand.
| Sweep | Total | MCP | LLM | OpenClaw | A2A |
|---|---|---|---|---|---|
| 243,615 | 1,997 | 20,026 | 192,492 | 32 | |
| 221,215 | 1,775 | 8,447 | 183,737 | 30 | |
| 2 sweeps · latest plotted | 204,866204,866 to 227,895 | 1,694 | 5,126 | 171,036 | 24 |
| 3 sweeps · latest plotted | 245,111245,111 to 349,424 | 1,807 | 47,369 | 153,638 | 25 |
| 2 sweeps · latest plotted | 297,723297,723 to 310,285 | 1,322 | 25,803 | 231,482 | 22 |
| 2 sweeps · latest plotted | 321,929313,900 to 321,929 | 1,134 | 25,128 | 267,172 | 8 |
| 294,277 | 889 | 23,086 | 242,379 | 6 |
Top countries
- United States57,247
- China36,817
- Singapore34,345
- Israel31,363
- Hong Kong SAR China11,003
- Germany9,716
- United Kingdom7,424
- Japan5,039
- Canada3,712
- Netherlands3,318
Top organizations
- ACEVILLE PTE.LTD.36,892
- Internet Rimon30,933
- Aliyun Computing Co., LTD11,217
- Tencent cloud computing (Beijing) Co., Ltd.10,305
- Fly.io, Inc.4,715
- Amazon Technologies Inc.4,393
- Amazon.com, Inc.4,328
- Avago Technologies U.S. Inc.4,120
- China Mobile Communications Corporation3,768
- Asia Pacific Network Information Center, Pty. Ltd.3,548
What this index does not do
- Does not access, test, or exploit any third-party system. All counts derive from publicly indexed search engine data.
- Does not estimate counts when sample size is insufficient. Empty categories are reported as zero, not extrapolated.
- Does not present a single composite ecosystem score. Each category is reported separately so changes in one do not obscure changes in another.