Biden administration unveils AI bill of rights laying out voluntary guidelines to avoid AI misuse

The White House released its artificial intelligence bill of rights Tuesday expressing the Biden administration’s vision for the future of private and public industry usage of the technology.

The "Blueprint for an AI Bill of Rights" (PDF) is an unenforced call to action laying out voluntary guidelines that companies developing or deploying AI can follow to safeguard people from misuse or abuse. Unless agencies or Congress enforce or expand legal protections, the guidelines will be followed on an opt-in basis.

Several federal agencies, including the Department of Health and Human Services (HHS), will be taking steps reflecting the guidelines’ goals, according to the White House. Topics covered range from protecting personal data to limiting surveillance. The 73-page document is partnered with a “technical companion” detailing steps to implement the guidelines.

“Automated technologies are driving remarkable innovations and shaping important decisions that impact people’s rights, opportunities and access. The Blueprint for an AI Bill of Rights is for everyone who interacts daily with these powerful technologies—and every person whose life has been altered by unaccountable algorithms,” said Office of Science and Technology Policy Deputy Director for Science and Society Alondra Nelson, Ph.D., in a press release. “The practices laid out in the Blueprint for an AI Bill of Rights aren’t just aspirational; they are achievable and urgently necessary to build technologies and a society that works for all of us.”

The Blueprint for an AI Bill of Rights addresses five broad principles.

The first guideline states that the American people have a right to safe and effective systems devoid of malicious intent or “reasonably foreseeable” threats to safety. Systems should “proactively protect” communities and individuals. This should be ensured by pre-deployment testing and risk identification, mitigation and monitoring.

The second “right” denotes that algorithms and systems should be equitable and free of discrimination. While the overall bill of rights has been criticized for being toothless, this guideline does make clear that in some circumstances, biased technology is illegal.

Rule three addresses data privacy by stating that users should have agency over how their data are used and should be protected from abusive data practices. “Any consent requests should be brief, be understandable in plain language and give you agency over data collection and the specific context of use; current hard-to-understand notice-and-choice practices for broad uses of data should be changed,” the guideline states.

The penultimate regulation asserts that users should know when automation is being used. These notices should be kept up to date, and those impacted should be informed when significant use or functionality changes.

Finally, users should be able to opt out of AI interaction and have access to a person who can promptly remedy issues when appropriate. This regulation may also be required by law in some circumstances including health, criminal justice and education.

“The Blueprint for an AI Bill of Rights and federal actions we are announcing today deliver on the President’s day one promise to support policies that advance equity and economic opportunity for the American people,” said White House Domestic Policy Advisor Susan Rice in a press release. “Taken together, these actions will help tackle algorithmic discrimination and address the harms of automated systems on underserved communities.”

The white paper reflects the administration’s stated agenda to hold technology companies accountable and ensure that new technological advances are equitable. It was formed after a yearlong consultation with more than two dozen departments, society groups, technologists, industry researchers and tech companies including Palantir and Microsoft.

While the bill of rights is nonbinding, the administration announced ways in which federal agencies will take actions to echo the white paper’s guidelines. Among them, HHS will issue “a vision for advancing Health Equity by Design” by the end of this year.

HHS released its “Trustworthy AI (TAI) Playbook” (PDF) last year to meet the Office of Management and Budget’s guidance for the regulation of the technology and Executive Order 13960, which encouraged agency AI innovation while defining relevant principles.  

“HHS has a significant role to play in strengthening American leadership in artificial intelligence,” said HHS Chief AI Officer Oki Mek in the playbook. “As we use AI to advance the health and wellbeing of the American people, we must maintain public trust by ensuring that our solutions are ethical, effective and secure. The HHS Trustworthy AI (TAI) Playbook is an initial step by the Office of the Chief AI Officer to support trustworthy AI development across the Department.”

The use of AI in a health space was mentioned throughout the playbook with specific guidelines to protect data privacy and avoid system or algorithm bias.

The solutions offered by AI diagnostic aids must be retraceable in the case of notable misdiagnoses. Healthcare research agencies should obtain and assess all documentation of security controls for third-party AI technology they employ. Transfer of private health information to third-party servers should be avoided. Back-end activity within electronic health record data should be closely monitored.

HHS suggests voice and speech systems should account for a broad range of languages; machine learning should be bias-free and incorporate relevant shifts in healthcare demographics.

Warnings toward common pitfalls include the use of healthcare spending in certain algorithms and predictive models, which HHS asserts is not a good indicator of future healthcare need. Models should also reflect the data of the populations in which they are being used.