The recently published White Paper of the European Union on the regulation of artificial intelligence (AI) contains guidelines to regulate AI so that “Europe becomes the most attractive, secure and dynamic agile data economy in the world. world”.
The strategy establishes a series of policy measures, including the mobilization of private and public investment , necessary to achieve this goal, and that prevent AI from “leading to violations of fundamental rights”, such as the introduction of biases in algorithms, suppression of dissent and lack of privacy.
To achieve this, among others, it suggests legal requirements such as:
- Make sure AI is trained with truly representative data, without racial, sexual or other biases.
- Require companies to maintain detailed documentation of how AI was developed.
- Conveniently inform citizens of when they are interacting with an AI.
- Require human supervision for AI systems . To guarantee the transparency of the system, the decisions they make, as well as their decision-making process, must be documented so that they can be understood, evaluated and explained by the responsible persons.
- Another element that raises concern among experts is the use of AI in lethal autonomous weapons systems. It is considered important that humans continue to be responsible for the casualties caused by such systems .
Criticism
The EU document is aimed at both developers and implementers and regulators of AI, with the purpose of being able to mark ethical boundaries of these systems, respect the current legal framework and promote democratic freedoms and the common good. However, the initiative has not been without criticism .
The new criteria appear to be much more flexible and vague than those suggested in an earlier version of the White Paper leaked in January. That draft proposed, for example, a moratorium on facial recognition algorithms in public spaces for five years, while this only requires a “broad European debate.”
It is usual that in the first drafts more demanding positions are raised as a political strategy, so there is little surprise given the withdrawal that has occurred in this final version. However, that is still disappointing just now that the problem of facial recognition in public spaces begins to cause greater problems to guarantee the privacy of citizens .
As if that weren’t enough, these EU guidelines only concern what are considered “high risk” technologies, which may include certain industries, such as healthcare, or certain types, such as biometric surveillance. But the suggestions would not apply to technology in the field of advertising nor would it affect consumer privacy in general terms, posing a risk to the freedom and autonomy of citizens .
All in all, this White Paper is just a set of guidelines. The European Commission will start drafting legislation based on these proposals and comments at the end of 2020.
