UnidosUS Applauds New Executive Order on Responsible AI Development and Calls for Further Protections

on

Washington, D.C. — UnidosUS, the nation’s largest Latino civil rights and advocacy organization, welcomes President Joe Biden’s new executive order directing federal agencies to advance responsible development of artificial intelligence (AI) in a manner that protects civil rights and safeguards privacy. UnidosUS President and CEO Janet Murguía issued the following statement: 

“President Biden’s Executive Order recognizes both the tremendous potential of AI as well as the real danger it poses. We applaud the Biden-Harris administration on its thoughtful leadership on AI and continuing efforts to promote fairness, transparency, privacy and accountability in tech. The order creates important new areas of focus across the federal agencies, producing principles and best practices to better serve the needs of patients, students and workers. 

“Latinos and communities of color have long endured bias and discrimination when new technologies emerge, from mortgage lending algorithms to social media targeting. We cannot allow AI to extend this shameful history. Together, we can harness innovation to expand freedom, equity and human potential. With thoughtful policies centered on fairness and accountability, AI’s development can be a new and different chapter in this story.” 

With more than 7 million new Latino workers poised to enter the workforce between 2020 and 2030, it is especially concerning that AI-enabled hiring tools have been shown to discriminate based on ethnicity. Latinos are also severely underrepresented in technology jobs, comprising just 3% of Silicon Valley “coders.” Sustained investments are needed to diversify both the opportunities and perspectives in the technology sector and its applications. 

UnidosUS looks forward to collaborating with the Biden-Harris administration and Congress to develop strong legal guardrails for AI that share its benefits while protecting Latinos’ civil rights. As our testimony highlighted, we also will need leadership in Congress and action from governments to keep pace with AI. 

For example, we should explicitly prohibit AI uses that enable discrimination and rights violations or pose a threat to our democracy, and we should require multi-stakeholder oversight processes to craft binding safeguards and push systems towards fairness. We should also fund timely and robust investments in AI ethics research and human infrastructure to equitably distribute its benefits and support accountability. Specifically, UnidosUS calls on policymakers to build on the executive order (EO) with the following additional steps: 

Protect Constitutional Rights and Civil Liberties: 

  • Ban government and private sector uses of AI systems for mass surveillance, predictive profiling, undisclosed algorithmic decision-making, facial recognition without a court order, social credit scoring systems and autonomous weapons. 
  • Require privacy by design principles that are compatible with effectiveness, such as strict data minimization, access controls and other privacy enhancing techniques for government AI uses. 
  • We applaud efforts to address algorithmic discrimination through training, technical assistance and coordination on civil rights violations related to AI, and the EO’s focus on developing best practices on AI across a wide range of criminal justice uses. We also support the call for federal agencies to strengthen privacy guidance and evaluate uses of commercial data but would urge specific legal prohibitions on rights-violating uses. 

Protect Election Integrity: 

  • Prohibit AI-enabled electoral mis- and dis–information, its use in voter suppression tactics, and the proliferation of unlabeled synthetic media that could impact electoral results. Require disclosures on AI-generated content. 
  • We applaud the EO’s steps to develop a watermarking and provenance system and urge the Federal Election Commission and Congress to take steps to address AI misuse in elections into law in a manner consistent with Constitutional protections. 

Enhance Transparency and Oversight: 

  • Mandate detailed documentation of government AI models, including publishing model cards, training data summaries, performance evaluations and explanations of purposes in both technical and plain language and create a consumer rating system on fairness, accountability, transparency and other factors.  
  • Require publicly reported independent audits by multi-stakeholder committees evaluating impacts of public sector AI systems on civil rights, privacy and equity.  
  • Allow individuals to request their data be removed from AI training datasets and require swift deletion after project completion absent a demonstrated legal purpose. 
  • While the EO’s safety and security provisions and creation of the AI Safety and Security Board are good starting points, we encourage publication of model cards and a consumer rating system for foundation models to facilitate public understanding.  

Require Regular Audits and Community Input: 

  • Mandate that both private sector and public sector entities undergo regular independent reviews to assess AI system safety, fairness, accountability and community impacts. Audits should gather input from diverse stakeholders and ethics professionals. 
  • The EO’s support for new guidance on AI auditing is a valuable step. We encourage building on this to require multi-stakeholder input and review for foundation models and high-risk government and industry uses. 

Protect and Support Children and Inclusion: 

  • Require expert evaluation and parental controls to protect children from risks of manipulative or inappropriate AI uses in educational and private settings. 
  • Encourage investments in educational tools for non-English learners and adaptive strategies for learners using AI for educational equity and inclusion.  
  • The EO calls for resources and guidance for AI educational tools. We encourage safeguards to address bias in evaluative uses of AI in educational advancement and safety rules for interactions with children developed through an inclusive and multi-stakeholder process.  

Invest in Ethical and Equitable AI: 

  • Fund research and community projects focused on accountable AI and fully and equitably sharing the benefits of AI, as we have proposed, through entities like an AI Ethics Institute modeled on the National Institutes for Health (NIH) and an AI Human Infrastructure Foundation modeled on the CDC Foundation. 
  • Support skills-based hiring and upskilling initiatives and efforts to increase diversity in employment pipelines for the technology sector.
     
  • Close the digital divide through long-term investments enabling low-income and rural Latino communities to access affordable broadband as well as digital skills programs to enable every community to fully participate in an AI-driven society.
     
  • We applaud the EO’s focus on AI’s impacts on workers and mitigating workplace harassment. Building on this with substantial, robust and timely investments in ethical research, digital access, tech sector diversity and community benefits would further advance an equitable AI future.