What cannot wait.
Critical AI safety patterns that need attention from everyone.
Not all research questions are equal. Some are intellectually interesting. Some are commercially valuable. And some are the difference between catching a dangerous pattern while it is still forming and catching it after it has hardened into the infrastructure that governs our lives.
This page lists the patterns where the cost of waiting is highest. The patterns where every month without focused research is a month the trajectory gets harder to reverse.
We are not asking you to fund us. We are asking you to pay attention. If you want to partner with us, the data is here. If you want to fund the research, we will put every dollar into the investigation. If you want to use our baseline data for your own work, take it. If you do not want to work with us at all, that is fine too. The data is still here. Come get it. Because these patterns do not wait for anyone's funding cycle, anyone's ego, or anyone's partnership terms.
How we classify urgency.
We classify research priorities on two axes, not one. Most labs only track threats. We track the full spectrum: what we must prevent and what we must capture.
The pattern exists now, the trajectory leads to serious harm at scale, the intervention window is narrowing, and insufficient research attention is being given.
The pattern is documented, the trajectory is concerning, research would prevent harm, and current attention is insufficient.
Early signals suggest this could become critical. Not enough data to confirm. Watching closely. Research and observation accelerate the answer.
Something unprecedented is appearing. A positive phenomenon, a new capability, a generative dynamic that has never existed before. Study it now or lose the window. The urgency is equal to the threat side. The polarity is reversed.
Classifications are reviewed quarterly as new data arrives.
Every item on this page is something we believe the world cannot afford to ignore.
We have the data. We have the methodology. We have the citizen science network that can detect these patterns at scale.
Bring your expertise. Use our baseline. We do not need to own the answer. We need the answer to exist.
Every dollar accelerates the timeline. These patterns are forming right now.
You do not want to work with us. Fine. The data is here. Take it. We would rather the problem get solved without us than not get solved at all.
You use AI every day. You see things no server log captures. Share the tool. Tell someone. The sensor network is people. It is you.