Trustworthy AI: Industry-Guided Tooling of the Methods

The need to assess and validate the trustworthiness of AI (robustness, transparency, safety, security, etc.,) has been the subject of considerable academic work for some time now. A natural evolution of such research efforts is to have a tangible impact in the industrial sector and in the upcoming standards. To this end, theoretical feasibility of algorithmic methods is not enough: one needs to put these methods inside usable tools that can scale to real-world problems. Evidently, this need has not gone unnoticed either and several teams are actively working on maturing their tools further and further in a constant race with a very rapidly moving field. While fundamental research is a paramount bedrock, in the present communication, we want to focus on how far we have come in satisfying the goal of seeing AI safely permeating our future. To this end, we will give a brief overview of recent collaborations with industrial actors in an effort to give the reader a wider notion of trustworthiness, one that may come into play on their own use-cases.