Calls for transparency in AI systems are increasing in number and urgency from diverse stakeholders ranging from regulators to researchers to users—with the notable (comparative) absence of companies developing AI. Many notions of transparency for AI have appeared, each addressing different interests and concerns.

In computer security, transparency is likewise seen as a key concept.The security community has for decades pushed back against so-called security by obscurity—the idea that hiding how a system works protects it from attack—against significant pressure from industry and other stakeholders. And over those decades, security researchers and practitioners have gradually built up some community norms around how to balance transparency interests with possible negative side effects.This paper asks: How can we leverage the experience (and mistakes) of the security community to inform the ongoing debate around transparency in AI?

We identify four key themes in the security community’s perspective on the benefits of transparency and their approach to balancing transparency against countervailing interests. For each theme, we investigate how each aspect yields insights relevant to transparency in AI. We also provide two case study discussions: One on the development of disclosure processes, a now well-accepted pillar of transparency in security, and the other on how transparency has shaped the research field of anonymization. Finally, shifting our focus from similarities to differences, we highlight key transparency issues where modern AI systems present challenges different from other kinds of security-critical systems, raising interesting open questions for the security and AI communities alike.