How good is the EU's plan for AI regulation?
The European Commission has presented the world's first legal framework governing the use of artificial intelligence (AI). In future, AI systems the use of which potentially pose a risk to the safety or rights of people will be subject to strict regulations. In the case of particularly clear risks, for example if free will is manipulated, a ban will apply. The initiative is aimed at boosting trust in AI.
Machines must learn values
Europe is assuming a pioneering role by setting ethical boundaries to protect democracy and privacy, La Repubblica says approvingly:
“The idea is to find a 'European path' between Chinese digital authoritarianism and American total market freedom. A culturally specific approach that Brussels has been preparing for years and that will be reflected in one of the key passages of the new rules: the algorithms on our continent must be 'fed' with data that reflects European values. ... If, on the other hand, it turns out that an artificial intelligence system is being fed with input that conceals prejudice, racism or discrimination, it must be reprogrammed from scratch.”
Include the private sector
Politiken welcomes the EU initiative, but would like to see even more restrictions:
“AI regulation, as envisioned by the EU Commission, seeks a third way between China's state-controlled surveillance capitalism and American laissez-faire. ... Among other things, it is supposed to stop state scoring of citizens, as China does, and limit police use of biometric mass surveillance. Hooray! But it can and should go further, not least to regulate the private sector, which is also enthusiastically using opaque algorithms.”
State use needs better regulation
For Der Standard, the EU's plans head in the right direction but make one annoying exception:
“Artificial intelligence is not neutral. It can't be, because in most cases it refers to machine-learning systems. ... For example, what information they have used to arrive at a result is known, and also what that result is. But no one really knows how this goal was reached. This opens the door to errors that are almost incomprehensible. However especially when it comes to investigating potential violations of the law, there must be no doubts. But in continuing to allow the use of AI for state surveillance in its draft, the Commission has not considered this aspect. ... So it has missed an opportunity to present an EU that is fit for the future and yet attaches great importance to fundamental rights.”
The game will be decided elsewhere
The decisive factor will be whether the EU succeeds in bringing its visions to the global level, comments the Süddeutsche Zeitung:
“Because the big game will be played out between the AI superpowers the US and China. ... Ultimately the EU should formulate its rules in such a forward-looking and comprehensible way that at least the US will want to abide by them. Under Biden's government, which is once again to looking to Europe, there is a real chance of this happening. A little counterforce against the lobbying might of the AI superpowers - Facebook, Google and Microsoft - won't hurt when it comes to shaping the invisible side of the future.”