The Council of Europe has called on its 47 member states to take a "precautionary" approach to the development and use of algorithmic systems so that they fully respect human rights.
A committee of ministers has issued a set of guidelines for governments to follow that recommends adopting legislation, policies and practices to ensure the systems they deploy do not violate human rights.
The recommendations acknowledge the vast potential of algorithmic systems to innovate industry processes and boost economic development, but it also points out the danger of bias and the undermining of human rights.
In 2018, Amazon had to scrap an AI-based system used for internal recruitment after it was found to be riddled with bias against female applicants. The programme was meant to filter out candidates into a top-five, but after a year of use, it became apparent it was failing to rate them in a gender-neutral way.
To stop systems like these becoming more common, the council is calling on their regulators to establish "effective and predicable" legislative and regulatory frameworks that prevent, detect, prohibit and remedy any human rights violations, whether that is from a public or private operator.
The areas the council is most concerned about are the challenges algorithmic systems present to the right to fair trials, privacy and data protection, freedom of thought and expression, equal treatment and also economic and social rights.
The IT expert’s guide to AI and content management
Your guide to the biggest opportunities for IT teams when it comes to AI and content management
FREE DOWNLOAD
The council believes that such is the impact of algorithmic development in these areas, member states must put in place effective risk-management mechanisms and also refuse to deploy systems that have the potential to lead to irreversible damage when they are "so opaque that human control and oversight becomes impractical".
In addition to the guidelines, the council is also encouraging governments to be in regular consultation and cooperation with relevant parties and that they foster general public awareness of the impacts of algorithmic systems, including their risks.
