Attention: You are using an outdated browser, device or you do not have the latest version of JavaScript downloaded and so this website may not work as expected. Please download the latest software or switch device to avoid further issues.
| 28 Aug 2024 | |
| AI |
At the core of the artificial intelligence (AI) revolution is data. Data are used to train machine learning models and shape the outputs of AI systems in ways that directly impact human lives and business models. To build safe and effective AI systems, the data that contribute to AI models must be high-quality and governed responsibility. Data policy is emerging as an essential discipline for governing how data are collected, shared, and used. Clear data policy principles and mechanisms for transparency and public accountability are necessary to uphold public interests and democratic values as the momentum behind AI continues to build. Without a comprehensive approach to data policy, AI has the potential to replicate and amplify existing issues reflected in the data rapidly and at scale.
Existing data and privacy laws like the Foundations for Evidence-Based Policymaking Act, Confidential Information Protection and Statistical Efficiency Act (CIPSEA), and the Privacy Act have been responsible for governing data in the past. However, the proliferation of accessible AI presents new challenges. AI requires vast amounts of data and can quickly replicate and magnify biases or errors present in that data on a large scale. Moreover, the diverse and often unpredictable future applications of AI create uncertainty about how to effectively protect data in this new landscape.
The rapidly-evolving technological environment poses significant challenges for policymakers and agency leaders. Not all policymakers can be expected to fully account for all aspects of data policy, nor identify each potential gap in data policies that will influence secure, trustworthy AI in the future. The Data Foundation aims to equip policymakers with a guide to understand what policies about data already exist, and identify potential gaps in existing laws or policies to help build a comprehensive approach to data for AI. To do so, this guide categorizes essential elements for responsible AI data use into three key components, and offers a tool to orient policymakers’ considerations related to data for AI, guiding the parameters for if, how, and what data should contribute to AI models. Key components of sound data policy outlined in this guide include:
There is no 'one-size-fits-all' approach to data policy in the age of AI. Ultimately, the determination of acceptable risks given potential harms and anticipated benefits is a policy decision that must balance the rapid advancement of AI technology with the need to uphold public interests and democratic values. Ideally, decisions about data for AI should not only be informed by a diverse set of experts, but also include input from stakeholder communities, including those potentially affected by AI systems and those represented in the data being used. A comprehensive approach to data policy is essential to ensure that as society becomes ready for AI at scale, the data contributing to AI models are also widely ready for responsible and ethical use.
The Data Foundation's guide will support assessments of whether potential uses, policies, and laws related to AI clearly articulate critical aspects related to sound data practices – and help policymakers identify areas where more legal clarity, guidance, or internal policy changes may be needed to ensure data are of high quality, governed by effective principles, and leveraged by qualified people using capable technical systems.
Suggested Citation: O’Toole, K., C. Turbes, and A. Freeman. (2024). Data Policy in the Age of AI: A Guide to Using Data for Artificial Intelligence. Washington, D.C.: Data Foundation.
DOI: https://doi.org/10.15868/socialsector.44213
