// AI Intelligence Hub · Wadsworth Digital

The clearinghouse
for AI intelligence.

Learn AI without the overwhelm. Try the major studios, follow teachers worth your time, and build from official training — updated when we refresh the data.

66,241 Exposed AI Systems

13 Studios & Links

— Learn Resources

$7.84B Agent Market 2026

Sep 16 Last Updated

⬡ Agents ◉ Exposure Tracker ⬢ Try Models ◎ Learn

Find the Right Platform

Not every AI agent is built for your use case. Here's how the major platforms stack up.

Platforms refreshed: 2026-09-16 · Exposure refreshed: 2026-09-16

By the Numbers

The market data behind the hype.

$7.84BAI agent market size in 2026

$52.6BProjected market by 2030

79%Enterprises with some agentic AI deployed

171%YoY growth in agent API calls (2024→2025)

5%Of exposed deployments with any auth enabled

66,241Exposed AI systems found in our last scan

Exposure Tracker

Exposure snapshot — AI infrastructure found on the public internet. Counts below are from the Sep 16, 2026 Shodan pass (first full refresh since Apr 9).

Last full scan: 2026-09-16 · Total exposed: 66,241 · Prior Apr 9 snapshot was ~79,800

Notable findings from Sep 16, 2026:

🔴 n8n — 24,581 exposed instances. Still the largest category by raw count, down from ~29K on Apr 9. Workflow automation with full execution capability remains the loudest public surface.

🟠 Open WebUI — 22,247 exposed instances. Up from the Apr snapshot on the continuity http.title query — full chat UIs tied to local models and external APIs, often with weak or no auth.

🟡 Ollama — 15,519 exposed instances. Down from ~19.8K on Apr 9. Still zero built-in auth on the default API — inference, prompt injection, and data exfil remain trivial when reachable.

🟢 Jupyter Notebooks — 3,883 instances. Down from ~5.5K in April. Browser-reachable remote code execution surface, quieter but not gone.

⚪ AutoGPT / LangChain (LangServe) — Near zero (10 / 1). Autonomous agent frameworks remain rare in the wild.

Our daily scan uses Shodan's API to query for known AI infrastructure signatures — service banners, default ports, and response fingerprints associated with Ollama, Jupyter, Open WebUI, n8n, and other agent platforms.

We do not interact with discovered systems beyond the initial handshake. No exploitation. No data collection from discovered instances. Scan is passive enumeration only.

Results are aggregated, anonymized, and published as trend data. Individual IPs are never published. Full methodology →

Try the models

Direct links to the big studios and open/local options. Click in, experiment, then come back to Learn for teachers and training.

Last updated: 2026-09-16

Learn AI

For newbs and builders. YouTube teachers, official training (Anthropic, OpenAI, Google, xAI), and tools. ⭐ = WD-recommended start.

Last updated: 2026-09-16

// Featured Analysis

The Hidden Attack Surface
Nobody Is Talking About

There are 66,241 exposed AI systems sitting on the open internet right now with zero authentication. Not test environments. Production deployments of Ollama, Open WebUI, n8n, and other agent infrastructure — reachable by anyone with a browser and a port scanner.

We know because we just scanned for them again on Sep 16, 2026 — first full pass since April. The mix shifted; the open-by-default pattern did not.

The organizations shipping the fastest are also the ones skipping the authentication step. They treat their AI infrastructure the way early web developers treated their databases — open by default, secured later. Except later never comes. And unlike a misconfigured MySQL instance, an exposed Ollama endpoint can execute arbitrary inference, exfiltrate training context, and be weaponized for prompt injection at scale.

The attack surface grows at exactly the same rate as the market. And the market is growing at 171% per year.

Run a Free Security Scan →

The clearinghousefor AI intelligence.