Anthropic mapped 832 AI-linked cyber threat cases
Anthropic studied 832 banned accounts tied to harmful cyber activity from March 2025 to March 2026. It found that attackers are using AI in more complex parts of attacks, not just for simple help. Solo makers using AI tools should treat account access, API keys, and automation settings as higher-risk areas.
Key points
- The report studied 832 banned accounts with enough detail for deeper review.
- Anthropic mapped the cases to MITRE ATT&CK, a shared guide to common attack methods.
- AI can help attackers connect several steps of an attack, which makes risk harder to judge.
- Check who can access your tools, where API keys are stored, and what automated actions are allowed.
Quick term guide
- Plex
- An app that streams your own video and music files to your devices.
- AI tools
- Software that can help create text, code, images, or other work.
- API keys
- Secret codes that let one app or service access another service.
- API key
- A private code that lets a service know which account is using it.
- automation
- A way to make repeated work happen without doing every step by hand.
- port
- A specific virtual door on your computer used by apps to send and receive information.
- MITRE ATT&CK
- A public guide that lists common methods used in cyberattacks.
- share
- A server folder made available to apps or other devices.