Desafíos actuales de la Inteligencia Artificial

A review of high-risk artificial intelligence (AI) systems that assess social security eligibility 253 Current social security benefits assessment 1 Country/ Region Incident Description Consequences United Kingdom Incident 189 Benefit fraud detection algorithm disproportion- ately targeted people with disabilities. DWP used ad- vanced AI to track possible fraud but refused to dis- close the algorithm. Facing fresh legal action to reveal the algorithms used. People with disabilities feel terrorized and anx- ious. Second legal letter sent by the same group on the subject, with the first going out before Christ- mas. United Kingdom Incident 611 Suspensions of Universal Credit benefits for many Bulgarian nationals. Poten- tial nationality-based tar- geting for benefit fraud in- vestigations led to poverty and homelessness. Equal- ity impact assessment identified disproportionate flagging of marriages from Greece, Albania, Bulgaria, and Romania. The UK government risks contempt of court unless it improves its response to transparency requests re- garding the use of AI for vetting welfare claims, ac- cording to the information commissioner. USA (Freep, 2023) $20 million settlement Class-action lawsuit filed by unemployment insur- ance claimants wrongfully accused of fraud by the MiDAS automated system. Claimants eligible for com- pensation if they received a fraud determination be- tween Oct. 1, 2013, and Aug. 31, 2015, and had assets seized on or after March 9, 2015. Settlement approved by the Michigan Court of Claims.