Europe has created a legal framework for artificial intelligence
This also concerns public administration.
Dr. S-W – 05/2021
The EC published a White Paper on
"Artificial Intelligence", COM(2020) 65 on February 19, 2020 The
European Representation of German Social Insurance (DSV) commented on this in
detail in a public consultation held on June 11, 2020, which can be found here.
The EC presented a regulation proposal about artificial intelligence as part of a package on April 21.
The basic rules are very detailed and they will also have far-reaching
implications for public administration.
AI might entail significant risks, not
least with regard to realising basic human rights and the principle of equal
treatment. Therefore, it is probably no coincidence that the controversial
"automatic facial recognition” complex is given a high priority in the
The Commission has further elaborated on
its "risk-based" regulatory approach already formulated in the White
Paper and it now distinguishes between four levels. The top two risk levels are
of particular interest from a social security perspective.
risk level concerns those AI applications, whose use is considered
legally/socially unacceptable and should be banned as a result. They mainly
refer to "social scoring". The wording is not very precise in detail
and it raises the question to what extent and for what purpose the creation of
personality profiles is still permissible.
However, the main
focus of the proposed regulation is on those applications that are considered
to be 'only' 'high risk'. They will continue to be permitted in the future, but
will be subject to a strict set of rules, including a number of documentation
requirements. They are intended to allow the so-called "market
surveillance authorities" to monitor compliance with the rules of the new
regulation. It must be taken into
consideration here that the same yardstick applied to European basic human
rights, including the prohibition of unjustified discrimination, will also be
strictly applied to AI applications. It is difficult to predict the outcome of
such mandatory checks.
The proposal is also based on a very broad
artificial intelligence concept It would probably cover many past programmes.
Indirectly, Thierry Breton, the European Commissioner for the Internal Market,
has confirmed this with the phrase "Although artificial intelligence has
been with us for decades, it has now reached new dimensions due to increased
The accompanying "European approach to
artificial intelligence", COM (2021) 205 final, notification that was
released on the same date, refers to the prominent role played by artificial
intelligence. Most recently, this had proven its potential in the fight against
COVID-19. The EU plans to invest 1 billion euros a year in this technology. 20
billion euros a year would also come from the private sector and the member
states, supported by the European Recovery and Development Plan. There is a close
link to the European data strategy, including the proposed Data Governance Act
and AI can only succeed if there is access to data.
As another part of the AI package, the
Commission revised the Coordinated Artificial Intelligence Plan adopted in
2018, see COM
(2021) 205 final. It calls on its member states to invest in AI, including
infrastructure and "data spaces". A European "centres of
excellence" network should support research and innovation. Finally, the
EU should lead the way in strategic sectors, including healthcare and the
public sector as a whole.