The Guiding Principles

The principles below reflect our organizational approach to the intersection of these technologies in our economy, with an emphasis on our two focus areas: housing and labor. 

Specifically, our principles are borne out of our existing research on tenant screening algorithms, the intersections of privacy, civil rights, and technology in the housing sector, the impact of automated management in the workplace on contract workers, and the rise of ghost work in the tech industry. Our recommendations are also informed by the incredible work of our organizational partners and experts in the field. In particular, we would like to acknowledge the work of the UC Berkeley Labor Center’s Technology and Work Program, AI NOW, Data & Society, ACLU, National Fair Housing Alliance, Economic Security Project, Ada Lovelace Institute, Electronic Frontier Foundation, Upturn, and Electronic Privacy Information Center (EPIC). 

Below we outline three major policy principles and accompanying recommendations to ensure responsible AI design, development, deployment, maintenance, and monitoring. We believe that a collective approach centering people in each step of the AI system—paired with a series of thoughtful, intentional choices now—will not only prevent the potential harms these tools could inflict but also create the conditions under which this promising new technology can enable human flourishing.

People who are impacted by AI must have agency to shape the technology that dictates their access to critical needs like employment, housing, and healthcare. 

Technology is most useful when it is designed based on real people’s needs. In that same spirit, the regulation of AI and digital technologies must take a human-centered approach that addresses current and potential harm, allows people to exercise their existing rights, and prioritizes accountability around AI tools as they are developed and deployed. These accountability structures should allow for meaningful, public control over the technology and its use in our communities, while at the same time allowing developers to experiment and innovate in ways that are guided by the input and expressed needs of the people who use and are impacted by the technology. 

Policy proposals that reflect this principle should:

1. Prohibit the use of unacceptable AI systems that are considered a threat to people (e.g. social scoring, biometric identification and categorization, etc.). 
2. Develop robust structures and standards that enact guardrails around the use of AI. 
3. Ensure clear roles for workers and communities in the design, development, deployment, and monitoring of AI systems. 
4. Require that an impacted (or likely to be impacted) person has the information they need to pursue or protect their existing rights.

The burden of proof must lie with developers, vendors, and deployers to demonstrate that their tools do not create harm—and regulators, as well as private individuals, should be empowered to hold them accountable.

Policy frameworks should create strong protocols for the review of tools before their deployment and in regular intervals post-deployment. Evaluating the extent to which tools are impacting people’s civil and human rights should not be a voluntary activity but rather required and enforced by regulators. Individuals should have access to information about how these tools are impacting their lives, but the onus of holding companies accountable should not fall solely at their feet. This will require transparency and disclosure on the part of AI developers, vendors, and deployers. Additionally, we must equip regulators and individuals with the capacity and resources they need to ensure that these tools do not threaten people’s safety or security, foster discrimination, or further systemic inequities in our economy. 

Policy proposals that reflect this principle should:

1. Require that developers, vendors, and deployers of AI systems affirmatively demonstrate that their tools do not create harm at each stage in the development, deployment, maintenance, and monitoring process. 
2. Strengthen requirements for oversight, transparency, and public reporting to address the opacity of these systems, their data, and their reasoning. 
3. Invest in growing enforcement capacity to adequately tackle the speed, scale, and technical complexity of AI and emerging technology.

Concentrated power and information asymmetries must be addressed in order to effectively regulate the technology.

Power imbalances in our economy enable exploitative applications of technology and AI. In order to effectively regulate the companies who are building and deploying it, these power imbalances must be addressed. This means stronger antitrust regulations and enforcement, increased attention to the incentive structures around the investment capital that enables the growth of the industry, stronger controls on the ability of companies to collect, store, and sell data, and a deeper understanding of the impact of the business models employed by companies building the technology. 

Policy proposals that reflect this principle should: 

1. Foster the conditions that reduce concentration through proactive guardrails that ensure competition in the AI market. 
2. Strengthen guardrails on the collection, sale, and monetization of proprietary data sets and hold data brokers accountable for their role in AI systems. 
3. Ensure publicly held, privacy-protecting data sets that can be used for research to strengthen our understanding of these technologies and their impact on our communities. 
4. Explore public-led innovation and create opportunities for public ownership to ensure that the growth of AI is equitable and widely beneficial to everyone.