Human rights in the age of

THE NEED FOR MORE RESEARCH OF FUTURE USES OF AI
While data protection, transparency, and accountability mechanisms go far toward mitigating human rights 
abuses in the use of AI, they don’t solve all of the foreseeable problems. For instance, as we identified, in the 
future, AI systems may substantially impact economic opportunities or facilitate war or conflict globally. For 
these reasons, we recommend that states and private-sector entities, including civil society organizations 
and individuals from academia, work together to investigate future uses of AI and continue to explore the 
potential human rights impacts identified here. Emphasis should be placed on identifying and building 
response mechanisms for potential threats to ensure that negative implications are mitigated to the fullest 
extent possible. These fora should be multi-stakeholder and pluralistic in order to ensure that all potential 
threats are identified and solutions don’t preference any specific group over another or further diminish 
marginalized voices.
123 UN Guiding Principles on Business and Human Rights, Principle 21.
124 For more on the procedural and substantive aspects of remedy in the information and communications technology sector, see the Access Now 
“Telco Remedy Plan” at https://www.accessnow.org/cms/assets/uploads/archive/docs/Telco_Remedy_Plan.pdf.
125 See Principles 29 and 31 of the UN Guiding Principles on Business and Human Rights for more information.

---

accessnow.org
36
HUMAN RIGHTS IN THE AGE OF ARTIFICIAL INTELLIGENCE
REBUTTAL: TRANSPARENCY AND EXPLAINABILITY WILL NOT KILL AI INNOVATION 
There are two frequently heard arguments against requirements for transparency and explainability of AI 
systems.
126
One argument is that AI is too complex to require transparency and doing so could damage 
innovation. The second argument is that explainability is impossible and, if mandated, would require AI 
developers to sacrifice the complexity of their systems and (again) hamper innovation. These arguments 
are overblown and inconsistent with developments in AI today. Each argument is addressed below.