On Thursday 4 December, the European Union Agency for Fundamental Rights (FRA) published a report setting out the difficulties involved in assessing the fundamental rights risks in the use of artificial intelligence.
The report, which is based on 38 interviews conducted in Germany, Ireland, the Netherlands, Spain and Sweden with suppliers, users and experts working in five areas considered to be sensitive under the AI Act – asylum, education, employment, law enforcement and social benefits – shows that while privacy, data protection and non-discrimination risks are often taken into account, other fundamental rights, such as the right to education or fair access to employment, are often overlooked.
In the FRA’s view, the definition of an artificial intelligence system should be broadened to include simple systems where the results, although sometimes biased, are not negligible.
The Agency is also calling for greater rigour in applying the exceptions provided for in the regulation, the so-called “filters”.
The FRA is also encouraging guidance intended for suppliers and users to carry out impact assessments on the full range of fundamental rights, as well as investment in independent research and testing to better understand the risks and how to mitigate them.
As Sirpa Rautio, Director of the FRA, noted in a press release, assessing these risks “is good for business” and helps to develop safer AI systems for citizens, as well as creating legal certainty for companies.
Read the report: https://aeur.eu/f/jv6 (Original version in French by Nithya Paquiry)