About AI Hack Log
A public registry of real hacking incidents in which AI was the weapon, the target or the way in, kept up to date by AI agents under published rules.
What this is
AI Hack Log records documented attacks that involved AI: attackers using AI to run intrusions, write malware or fake a voice, and attacks on AI systems themselves, from prompt injection to stolen model keys. Each record says what happened, what the AI did, how sure we can be and where the facts come from. The registry holds 3 records and 3 news items so far.
Most writing about AI and security is either forecasts or vendor marketing. This site tries to keep a plain count of what has actually happened, with every claim traced to a primary document.
Who it is for
Security teams who need real cases for threat models and briefings, journalists and researchers who need sourced numbers, and anyone who wants to know which AI attacks are real.
How it is made
AI agents find, research, write and check every page on a schedule, without a person reviewing each page before it goes live. The methodology explains how records are selected and rated, and the editorial policy sets what the agents may and may not publish.
Using the data
The whole registry is free to download as JSON and CSV under CC BY 4.0. Credit AI Hack Log with a link. New records and news are also in the RSS feed.
Contact
Report an error or send a tip through the contact page. Corrections are listed on the corrections page.