EU
AI Act : Summary and Analysis
-Rishiraj Chandan
In
this blog, we'll talk about the EU Artificial Intelligence Act, the first ever
comprehensive attempt at regulating the uses and risks of this emerging
technology.
Introduction
Artificial
intelligence is here to stay, and these technologies are expected to bring a
wide array of economic and societal benefits to a wide range of sectors,
including environment and health, the public sector, finance, transport, home
affairs, and agriculture. Already, AI is used to improve health care, optimize
service delivery, efficiently manage energy consumption, and drive human
progress in countless ways. Companies use AI-based applications to optimize
their operations, but the deployment and use of AI entails both benefits and
risks that trigger major legal, regulatory, and ethical debates. Indeed, the
main concerns are privacy, bias, discrimination, safety, and security.
Given
the fast development of these technologies, finding a balanced approach to
regulating AI has become a central policy question in the EU, and the key issue
is how to minimize risks and protect users without curtailing innovation and
the uptake of AI. After many debates, studies, and impact assessments, the
commission thought it had found the answer: the Artificial Intelligence Act.
It's the first ever attempt to enact a horizontal regulation of AI, and the EU
wants it to be an example for the rest of the world to follow. Now, what are
the main points of this regulation?
Objective
The
main objective is to lay down a common legal framework for the development,
marketing, and use of AI products and services in the EU. Through this act the
policy makers aim to make Europe world-class in the development of secure,
trustworthy, and human-centred artificial intelligence, and of course, in the
use of it. EU regulation addresses the human and societal risks associated with
specific uses of AI, and this is to create trust. On the other hand, act’s
coordinated plan outlines the necessary steps for member states that they
should take to boost investment and boost innovation. All this to ensure that
European member states strengthen the uptake of artificial intelligence in
Europe. The new framework lays down different requirements and obligations for
the development market's placement and use of AI systems based on a risk-based
approach.
According
to this pyramidal approach, AI systems presenting a clear threat to people's
safety and fundamental rights would be banned from the EU market due to the
unacceptable risks they pose. Talking about, for example, systems that exploit
vulnerable groups or AI systems used by public authorities for social scoring
purposes. A wide range of high-risk AI systems would be authorized for
commercialization and use, but subject to a set of requirements and
obligations, particularly on conformity, risk management, testing, data use,
transparency, human oversight, and cyber security. AI systems presenting only
limited risk, such as chat bots or biometric categorization systems, would only
need to comply with basic transparency obligations. All other AI systems
presenting only low or minimal risk could be developed and used in the EU
without additional legal obligations. What could happen if similar technologies
fall into different categories depending on their usage?
Let's take facial recognition as an example.
Facial recognition systems that are increasingly used to identify people can be
very useful to ensure public safety and security but they also can be
intrusive. The risk of algorithms making errors is high, so the use of those
technologies can affect citizens fundamental rights and result in
discrimination, in violation of our right to privacy. Even lead to mass
surveillance and that's why the commission's artificial intelligence act wants
to differentiate these systems according to their high risk or low risk usage.
For instance, using real-time facial recognition systems in public places for
law enforcement purposes would be prohibited with some exceptions because of
the significant threat to fundamental rights. But a wide range of those
technologies used for controlling borders or the access to public transports
and supermarkets could still be a load subject to strict controls, safety
requirements. Other of these technologies would be considered to have limited
or low risk and therefore would be subject to even less stringent rules. But
the difference between low risk and high risk usage is not always easy to make.
So, how's the system going to work in practice?
There
will be significant impacts particularly for providers of high risk AI systems.
They would have to comply with a range of requirements to protect users safety,
health and fundamental rights to be allowed to sell their AI products and
services in the union. National market surveillance authorities would be
responsible for ensuring that operators comply with the obligations and
requirements for high-risk AI systems and to restrict or withdraw them from the
market if they fail to do so. EU artificial intelligence board would be set up
to facilitate the implementation of the new rules and ensure cooperation
between the national supervisory authorities and the commission. However,
experts warn that EU must first get the AI systems definition right.
Conclusion
All
in all, I think that the proposal of the European Commission for an AI Act is
well designed. However, what is problematic is that the definition of what is
an AI system is very broad. It encompasses not only machine learning systems
but potentially all kinds of software and therefore might lead to
over-regulation. This is not the only concern. Another problematic point of the
proposal is the enforcement because the proposal basically relies on high-risk
AI systems based on the self-assessment of providers, and that is, of course, a
very weak enforcement structure. There are no individual rights of individual
citizens; there are no collective rights of citizen rights movements and
consumer organizations; therefore, there is the danger that the regulation once
it is enacted will not be enforced, and criticism is prevalent not just by one
or two experts but many. Other experts and stakeholders are also calling for a
number of amendments, including narrowing down the definition of AI systems to
focus on those that could really pose a risk. Broadening the list of prohibited
AI systems and ensuring proper democratic oversight of the design and
implementation of European regulation of AI Member states have been reviewing
the proposal for some time, and the European Parliament is also busy with it,
but there's still a lot of work ahead if EU want to make sure the Artificial
Intelligence Act will achieve its twin aims of ensuring safety and respect for
fundamental rights while stimulating the development and uptake of AI-based
technology in all sectors, and it's an enormously important task. The AI Act is
the first of its kind anywhere in the world, and it may set global standards
for the deployment of AI systems. It will regulate an incredibly broad set of
technologies and tools across all sectors, so it's important to get it right.
No comments:
Post a Comment