Working with NOPE
If you run an AI product where people talk about hard things — a companion, an assistant, a platform with a safety team — there are more ways to work with us than metering an API.
We publish our models, benchmarks, and incident research in the open so every platform — including ones that never pay us — handles these conversations well. Partnership starts from that posture.
Run our models inside your stack
Edge, our fine-tuned crisis-detection classifier, is distributed as open weights under MIT. Ocular OSS, our behavioral classifier, is Apache-2.0. Both are free for any use, including commercial, with no conversation data leaving your environment.
If you want more than the weights, we offer packaged on-prem builds, calibration against your domain, and support contracts — delivered through our customer portal, on your infrastructure.
Already running Edge or Ocular OSS?
Talk to us — no strings. The licenses don't require it, and nothing changes if you don't. But we'd rather help you deploy well than leave you guessing: operating thresholds, calibration quirks, known failure modes, what's coming in the next release. We don't ask who you are before answering.
Co-development
Detection calibrated on scripted test cases only goes so far. We work with a small number of design partners to calibrate detection on their real conversation domains — their personas, their user populations, their failure modes — and both sides keep what's learned: you get detection tuned to your product; the open models and public taxonomies get better for everyone.
Our benchmarks stay independent
One thing partnership never buys: a different number. Results on our public test suites and open benchmarks are not affected by commercial relationships, and we don't accept payment to re-run or improve a score. If that independence is why you'd trust us, it's also why it's not negotiable.
Start a conversation
Tell us what you're building and where safety fits. A short email is plenty.