Human rights in the age of

1. Follow open procurement standards.
When a government body seeks to acquire an AI system or components 
thereof, procurement should be done openly and transparently according to open procurement standards. This 
includes publication of the purpose of the system, goals, parameters, and other information to facilitate 
public understanding. Procurement should include a period for public comment, and states should reach out to 
potentially affected groups where relevant to ensure an opportunity to input.
2. Mandate human rights impact assessments. 
States must thoroughly investigate AI systems to identify human 
rights risks prior to development or acquisition, as well as on a regular and ongoing basis throughout the lifecycle 
of the system. A human rights impact assessment may be a necessary part of a larger algorithmic impact 
assessment process that examines broader threats, including threats posed by uses of AI to conduct 
surveillance or other activity that interferes with human rights. 
117
Appropriate laws must exist to govern 
116 See https://www.ohchr.org/EN/ProfessionalInterest/Pages/InternationalLaw.aspx for a summary of state’ human rights obligations under 
international law.
117 The AI Now Institute has outlined a practical framework for algorithmic impact assessments by public agencies, https://ainowinstitute.org/
aiareport2018.pdf. Article 35 of the EU’s General Data Protection Regulation (GDPR) sets out a requirement to carry out a Data Protection Impact 
Assessment (DPIA); in addition, Article 25 of the GDPR requires data protection principles to be applied by design and by default from the conception 
phase of a product, service or service and through its lifecycle.

---

accessnow.org
33
HUMAN RIGHTS IN THE AGE OF ARTIFICIAL INTELLIGENCE
the uses of AI for these purposes.
118
Any assessment process should include: 
• Testing and audits by independent experts 
• Identifying measures to mitigate identified risks and prevent any rights violations from occurring, 
and measuring compliance and efficacy 
• A failsafe to terminate acquisition, deployment, or any continued use if at any point an identified 
human rights violation is too high or unable to be mitigated 
• Identification of any new legal safeguards needed to protect human rights in specific applications of 
AI tools
• Special determinations of bias, particularly in the criminal justice sector due to the risks to fair trial, 
right to liberty, and non-discrimination
• If a third party is used to develop and/or implement the system, a requirement for the third party to