Led by noted technologist Tim O’Reilly and economist Ilan Strauss, the AI Disclosures Project focuses on addressing the potentially dangerous consequences for society’s safety and equity that might arise from how AI is commercialized. Through enhancing corporate transparency and disclosures, the AI Disclosures Project aims to ensure that economic incentives don’t lead corporations to ignore risks or otherwise harm users of their AI technology products.
Current AI governance frameworks focus on risks that are inherent in the capabilities of the models themselves, on limiting the ability of various bad actors to use them for harm, and on the security of various high-risk domains. These frameworks don’t adequately consider how AI risks may also originate in how companies compete for market share and profits. Companies may “move fast and break things” in order to gain scale while the market is still young, and they may exploit their market power once AI markets mature. In addition, they may successfully identify AI risks but under-invest in countering them. Furthermore, as the power of AI models grows, the risk profile may change, and current safety practices may be insufficient.
Alignment of AI products with societal objectives is impossible without corporate disclosure and auditing of AI-related risks. As Tim O’Reilly notes: “If we want prosocial outcomes, we need to design and report on the metrics that explicitly aim for those outcomes and measure the extent to which they have been achieved.”
Through high-quality research, collaboration, and policy engagements, we aim to develop a systematic disclosure and auditing framework that can become the basis for a set of “Generally Accepted AI Management Principles.” We will work to first understand companies’ AI-related business practices and the metrics they use to measure and manage the systems they build. Our goal is to learn from companies that are acting responsibly, and to use their best practices to shape disclosure standards for AI auditing and regulation that are informed by the commercial realities of AI markets.
The AI Disclosures Project is funded through generous grants from the Omidyar Network, Alfred P. Sloan Foundation, and the Patrick J. McGovern Foundation.