Why AI Threat Watch exists
Artificial intelligence is one of the most useful technologies of our lifetime. Like every powerful technology before it, it is also being turned against people. This site exists so that nobody has to learn that the hard way.
The harm is real and it is happening now
Voice clones empty bank accounts. Deepfakes move elections and markets. Autonomous agents break into government portals and company networks in hours. Poisoned packages ride into AI toolchains and steal the keys to the cloud. None of this is theoretical: every case file on this site is a documented incident with named sources, and the wire adds new reports every 30 minutes.
The record was scattered, so every victim started from zero
Until now the evidence lived in vendor blogs, court filings, agency advisories and local press in a dozen languages. A bank in Jordan, a hospital in Spain and a newsroom in Brazil could each be hit by the same technique and never know the others existed. A threat that is already documented should never surprise anyone twice.
What we believe
The benefits of AI are worth defending, and defending them starts with honesty about the harm. We name incidents precisely, we separate what is verified from what is merely reported, and we learn from every case the way aviation learns from every crash. Fear is not the goal. Preparation is.
What we do about it
We gather every publicly reported AI security incident, attacks on AI systems and attacks carried out with AI, in near real time from more than fifty vetted sources. Each one is verified and filed as a permanent case file with its sources, severity, timeline and attribution, published in English, French, Spanish, Arabic and Chinese, and released as open data under CC BY 4.0 so anyone can build on it.
Who it is for
Journalists who need a verifiable record before they publish. Security teams who want to know which technique is being used against organisations like theirs this week. Finance and risk executives who have to price the exposure. Researchers and policymakers who need a dataset rather than anecdotes. And every community that deserves a warning before the disaster, not an explanation after it.
Who is behind it
AI Threat Watch is an independent project started in 2026 by Karl Abou Zeid. It is independent: no company pays for a case file, a rating or a mention, and no incident is left out or softened for money. The research, triage and translations are done with AI tools and checked by a human editor; we say so because a site about the misuse of AI should be transparent about its own use of it.
How you can help
Report an incident we have missed, correct one we got wrong, subscribe to the alerts, reuse the data, or tell someone who needs it. Every contribution makes the next attack a little less likely to succeed.
Report an incident Get alerts Method, licence and corrections