Launching the Really Simple AI Risk Framework

We’re excited to announce the launch of the Really Simple AI Risk Framework, your go to tool which helps you identify and mitigate AI/ML-specific risks before they become incidents.

Why we built this

AI products fail at specific points: a poisoned training set, an unmonitored production deployment, a prompt injection nobody caught, or they might just end up having critical unintended consequences in the form of harmful side-effects the model produces without explicit intent or anticipation by its makers. Most existing risk frameworks are either too broad, or too complex to adopt quickly and/or effectively. Our aim with the Really Simple AI Risk Framework is to close that gap & offer AI product owners and deployers a curated, plain-language risk register that teams can actually use, built from industry insights and real incidents.

Unique Features

We built the framework in a comprehensive & easy to use manner, around the AI life-cycle stages: Development, Deployment, Use, and Monitoring since not every team builds models, but every AI initiative touches deployment and use, and monitoring gaps are some of the most common places risk hides.

One of its superpowers, besides it being really simple, is that it is a living, community-fed framework. If you spot a risk you do not see catalogued, you can propose it directly via an issue on the GitHub repository, helping everyone else benefit from what you have already discovered.

Photo by Canva