Fifteen associations are taking the lead and petitioning the Council of State to obtain the deletion of an algorithm used by the National Family Allowance Fund (CNAF). This algorithm is used in particular to better detect fraud and undue payments among its beneficiaries.
In detail, the appeal, filed Tuesday evening with the authority, “concerns both the extent of the surveillance at work and the discrimination carried out by this algorithm towards beneficiaries who are already vulnerable in their life courses », argue in a press release Amnesty International, Quadrature du Net, the Abbé Pierre Foundation and the other requesting associations.
Social Security: a massive savings plan on healthcare reimbursements
“This algorithm assigns to each recipient a suspicion score, the value of which is used to select those subject to inspection. The higher it is, the greater the probability of being checked “, they criticize, specifying that this tool “ analyzes the personal data of more than 32 million people living in a household receiving a CAF benefit ».
Overcontrolled people in difficulty
After having access to the “ code source » of a version of this algorithm, used between 2014 and 2018, the associations affirm that “ among the factors that increase a suspicion score, we find in particular the fact of having low income, being unemployed, benefiting from active solidarity income (RSA) or the disabled adult allowance (AAH) ».
« In return, people in difficulty find themselves over-controlled compared to the rest of the population. », they denounce.
In July, the associations asked the CNAF to stop using this algorithm: “ To the extent that, after two months, we did not receive any responses from the CNAF, this gave rise to an implicit decision of refusal. », Explains Katia Roux, Technology and Human Rights advocacy officer at Amnesty, to AFP.
Health: why mutual prices could rise by 3%
An algorithm to spot errors
Used since 2011, the statistical tool aims to identify among the 13.5 million beneficiaries most likely to make errors in their declaration, the director general of the CNAF, Nicolas Grivel, indicated in November 2023.
Beneficiaries of certain social minimums, such as the RSA or the activity bonus, whose income often varies, must complete quarterly income tax returns with complex forms and are more at risk of making errors. According to the CNAF, this algorithm aims to identify these beneficiaries to carry out checks quickly and rectify errors.
« Our objective is to limit as much as possible declarative errors and their consequences in terms of overdue generation. », assured Nicolas Grivel, affirming that it is “not not discriminatory » and does not target “ not necessarily the poorest people but those whose income varies ».
The Social Security deficit passed over in silence by Michel Barnier
Costly mistakes
In May, the Court of Auditors singled out the accounts of the family branch of Social Security, estimating it at 5.5 billion euros. the amount of errors not corrected by internal control actions ». Et « errors related to the data taken into account to pay benefits » represent “ 7.4% of the amount of benefits », indicated the Court in its annual report on the Social Security accounts. They concern in particular the RSA, the activity bonus and housing assistance. “ In particular, a quarter of the amounts paid under the RSA are tainted by errors », she pointed out.
« The network’s error detection capacity remains much lower than the risk induced by the insufficient reliability of the data declared by beneficiaries. “, she added.
However, the establishment of the “ solidarity at the source ”, from the end of 2024, should make it possible to “ significantly reduce the risk » of errors, Nicolas Grivel told AFP, since the declarations will be pre-filled.
The path to compulsory insurance to finance nursing homes
Furthermore, the family branch carried out a fraud assessment in 2023, as it does every two years. According to the Court of Auditors, potential fraud “represents 4.9% of legal benefits paid » by Cnaf, or 3.90 billion euros, up 39% compared to the previous assessment in 2021.
(With AFP)
