How did it all start?
The European Union is discussing the possibility of banning the use of artificial
As part of the draft resolution“On a European Approach to Artificial Intelligence,” the EU executive proposes to ban AI technologies that are used for “mass surveillance applied in a generalized manner to all individuals without any distinction.” The text of the bill describes in detail such surveillance methods as “monitoring and tracking of individuals in the digital or physical environment, as well as automatic aggregation and analysis of personal data from various sources,” reports REGNUM news agency.
The European Commission seeks to declare outof the law of artificial intelligence (AI) systems used for "mass surveillance of people" as part of new bans that will be introduced next week.
In addition, the European Commission provides for bansto use AI applications that violate EU values or violate human rights. These include systems that manipulate human behavior and target vulnerable populations.
According to the draft, EU members, for example, will have tocreate evaluation committees to test and validate high-risk AI systems. And companies that develop or sell AI that are subject to restrictions, including those from other countries of the world, can be fined up to 4% of their global income.
What is “high-risk” artificial intelligence?
According to the planned rules, those whointends to apply artificial intelligence, will have to determine whether a particular use case is "high risk" and therefore whether they need to conduct a mandatory pre-conformity assessment or not.
Examples of high risk AI:
- systems that prioritize the dispatching of emergency services,
- systems that determine the access or appointment of people to educational institutions,
- recruitment algorithms,
- those that assess creditworthiness,
- to carry out individual risk assessments,
- algorithms for predicting crimes.
What is the essence of the AI control bill?
We explain what the new bill provides.
Collection of data about citizens using systemssurveillance, monitoring of social networks, banking information and other sources - “indiscriminate surveillance of individuals should be prohibited if it is applied in a generalized manner to all individuals without distinction.” In other words, “selective” surveillance, if an individual has something of interest to the competent authorities, may well be allowed, but total surveillance “just in case” will probably be prohibited in Europe.
The bill also provides for a ban onintroduction of a social rating or social crediting system, when an array of big data on connections, behavior, purchases and other information is accumulated on citizens, and a social portrait of the citizen is created based on the totality of data. AI or neural networks can also help figure this out. They love this in the banking sector, and Europe is no exception in this regard. But China has gone the furthest in creating a social credit system, where people with low ratings are denied tickets for intercity transport, residence in certain areas, and a number of other restrictions are introduced. Such AI systems have a “high chance of influencing someone’s livelihood,” the paper notes.
Regarding the social rating in the billsaid: "AI systems used to assess social indicators of general purpose should be banned." Voting on this bill is expected on April 21. Until this point, representatives of the European Commission refuse to comment on the unapproved document.
The project also provides for:
- notifications about human interaction with AI, unless it is “obvious from the circumstances and context of use”,
- assessment of high-risk systems before they arecommissioning, including ensuring that these systems are understandable to human supervisors and that they are trained on "high quality" data sets tested for bias,
- creation of a “European Council on Artificial Intelligence”, consisting of representatives of all EU countries.
What do experts think about AI control?
Perhaps the most important section of the documentis an article that bans certain uses of AI, including mass surveillance and social credit ratings. However, reactions to the project from digital rights groups and policy experts say this section needs to be improved.
“Descriptions of prohibited artificial systemsintelligence are vague and contain unclear language that can create serious loopholes,” explains Daniel Leufer, European policy analyst at Access Now, in an interview with The Verge. This section, in his words, is “far from ideal.”
How exactly does national legislation decideDid the AI decision cause harm to anyone or not? On the other hand, the ban on the use of AI for mass surveillance is “too lenient.” He adds that the ban on social credit systems based on "reliability" is also defined too narrowly. Social credit systems do not need to assess whether someone is trustworthy to decide issues such as their eligibility for benefits.
At Twitter Omer Tene, Vice Presidentnon-profit organization IAPP (International Association of Privacy Professionals), commented that the ruling “represents a typical Brussels approach to new technologies and innovation. When in doubt, adjust." According to Tene, if the proposals are adopted, it will create a “vast regulatory ecosystem” that will attract not only creators of AI systems, but also importers, distributors and users, and will create a number of regulatory boards, both national and ;across the EU.
However, this ecosystem will not be the firstIt’s the turn to limit “Big Tech,” says Michael Veale, a lecturer in digital rights and regulations at University College London. “His focus is primarily on lesser-known business and decision-making tools providers whose work often eludes the scrutiny of regulators or their own clients.”
It is unknown what changes may have been madein this draft proposal as EU policymakers prepare for an official announcement on April 21 . However, once a regulation is proposed, it will be subject to changes following feedback from MEPs and will need to be implemented separately in each nation state.
Why is a new law needed?
A plan to regulate AI has already been discussed for some time.time. Back in February 2020, the European Commission published a white paper outlining plans to regulate “high-risk” artificial intelligence applications. Although it is not entirely clear from the new draft how exactly “high risk” will be determined.
According to the draft draft, the main goalThe Commission's goal here is to increase public confidence in AI through checks and balances based on EU values. The goal is to stimulate the adoption of so-called “trustworthy” and “human-centered” AI. This applies not only to “high-risk” applications, but to everyone that interacts with a person.
What will happen if the law is passed?
If the draft proposal is accepted, it willwill result in the EU taking a strong position on certain AI applications. Some uses will be subject to controls similar to EU digital privacy regulation under GDPR legislation.
EU requirements include ensuringsafety and ensuring the consistent accuracy of the AI —with the condition of reporting “any serious incidents or any failure in the operation of the AI system that constitutes a breach of obligations to the supervisory authority no later than 15 days after became aware of this.
It is noteworthy that the commission avoidedproposals to ban the use of facial recognition in public places. According to a leaked draft early last year, she was considering the idea.
It is not known what changes might have been made tothis is a draft proposal as EU politicians prepare for the official announcement on April 21st. However, once the regulation is proposed, it will be subject to change after receiving feedback from MEPs and must be implemented separately in each nation state.
This is not the first time Europe is talking about regulationtechnology industry. Thus, the EU authorities want to force Google and Facebook to pay publishers for news, and at the beginning of the year, European politicians called the heads of tech giants to discuss laws limiting their monopoly. By all accounts, the laws being drafted will provide for a 10% revenue penalty for tech giants to limit their monopoly.
Read more
The first accurate map of the world was created. What's wrong with everyone else?
NASA told how they will deliver samples of Mars to Earth
An engine for an orbital aircraft was tested in Russia
General Data Protection Regulation, GeneralThe General Data Protection Regulation (GDPR) is a regulation of the European Union through which the European Parliament, the Council of the European Union and the European Commission strengthen and unify the protection of personal data of all individuals in the European Union ( EU). The regulation also aims to export data from the EU.
The GDPR is aimed primarily at ensuring thatto give citizens control over their own personal data, and to simplify the regulatory framework for international economic relations by unifying regulation within the EU.
The social credit system ("the socialrating "," system of social trust ") - a system for evaluating individual citizens or organizations by various parameters, the values of which are obtained using mass surveillance tools and using big data analysis technology. It was first implemented by the Chinese government in the 2010s.