Transparency, in the AI of the Beholder

A leading tech nonprofit and think tank has issued a “Global Declaration On Free and Open AI” amid this week’s various AI summits on the Hill.

The statement from the Information Technology and Innovation Foundation (ITIF), signed by more than a dozen groups, is notable in flipping the traditional definition of “transparency” in AI on its head. Rather than explicitly advocate for the “open” model of AI development over “closed” models, ITIF makes a pro-business argument that both philosophies can help promote democratic values around the world — provided governments don’t get in the way.

“Transparency requires governments to avoid manipulating AI models, such as by prohibiting responses to certain legitimate topics, requiring models to produce certain responses, or intentionally introducing bias into datasets used to train AI models,” they write. “Transparency promotes both accountability and trust and is important to prevent governments from distorting AI-enabled speech and information.”

The statement urges nations to commit to principles of free expression and the free flow of information, among other tenets of “free and open AI.”