We Scanned 1 Million Exposed AI Services. Here's How Bad the Security Actually Is
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Researchers scanned 1 million publicly accessible AI services and found widespread security flaws
- Businesses are moving fast to self‑host LLM infrastructure, driven by AI productivity pressure
- Security gaps such as default credentials and exposed APIs were common across the scanned services
Why it matters: Enterprises that self‑host LLMs risk data breaches, ransomware attacks, and costly downtime, as the scan uncovered 1 million exposed services with weak authentication, open APIs, and default credentials.
