The Senator intervenes in the European Commission. It is about bots, AI and the safety of the European information space
Senator Adam Szejnfeld asked the European Commission to take action on the increasing threat of automated online impact operations. The focus was on bots, generic artificial intelligence, deepfakes, and coordinated networks of non-authentic accounts that can be utilized to manipulate public debate and artificially form the coverage of circumstantial messages.
Intervention not only concerns current threats, but besides questions as to whether the regulations in force in the European Union are keeping pace with the fast improvement of technology.
The Senator calls for the European Commission to carry out a comprehensive analysis of the existing rules, prepare common method standards for online platforms and consider further legislative changes.
New technologies – immense opportunities but besides fresh threats
The improvement of artificial intelligence has changed the way information is created and distributed online. Language models, image generators, audio and video tools, and automatic publication systems present make it possible to produce immense amounts of content in time, which was not available until a fewer years ago.
These technologies, of course, have tremendous possible and have legal application in administration, education, the economy, wellness and science.
The problem arises erstwhile the same tools are utilized to conduct coordinated impact operations.
Artificial intelligence can not only be utilized to make a single fake material. In combination with the automation of publications and a large number of online accounts, it can enable mass action.
This may mean, among another things, artificially expanding the coverage of certain narratives, creating apparent public support, manipulating trends, disrupting the operation of recommending systems or mass distribution of AI-generated materials.
As the Senator points out, the problem is becoming more and more so not only the content published on the Internet, but besides how circumstantial content is distributed and the network of accounts involved.
DSA and AI Act – an crucial step but enough?
The European Union has already created an crucial legal framework for the functioning of the digital world.
The Digital Services Act (DSA) Regulation introduced fresh obligations for the largest online platforms, in peculiar in terms of identifying and reducing systemic risks. These concern, inter alia, the risks of manipulating services and the negative impact on public debate.
In turn, the AI Act introduces certain transparency requirements for materials generated or manipulated utilizing artificial intelligence.
According to the Senator, however, the problem is that technological improvement is highly dynamic.
Mechanisms that have yet to appear adequate may not correspond to the scale and nature of the fresh threats.
It is peculiarly crucial that modern impact operations are increasingly not active in publishing 1 fake material. Their essence is the usage of a full network of accounts, automation and advice algorithms.
The problem is not only false content
One of the most crucial elements of the Senator's proposal is to draw attention to the phenomenon referred to as Coordinated Inauthentic Behaviour – CIB, or coordinated non-authentic behavior.
In practice, it is simply a situation in which many accounts operate in a coordinated manner to produce a circumstantial effect.
This can be e.g. artificial increases in the number of views, likes, comments or sharing. It may besides be simultaneous publication of akin content by a large number of accounts or the usage of automation to influence trends and recommendations systems.
In this case, an analysis of a single alert may not be sufficient.
It is important, first and foremost the behaviour pattern of the full network.
Therefore, the paper addressed to the European Commission points out the request to make methods of analysis specified as synchronisation of publications, similarity of content, links between accounts, characteristics of devices and method infrastructure, or repetition of certain behaviours.
European Bot recognition Standard
One of the demands is to make a common European model for identifying accounts utilizing automation.
Currently, individual platforms can usage their own definitions and their own bot detection systems. This makes it hard to compare data and conduct analyses covering respective platforms at the same time.
The proposed solution would imply defining a minimum set of information on accounts utilizing automation.
These could include:
- the level of account activity automation,
- the usage of a generic AI,
- the kind of software used,
- the way content is published,
- the minute erstwhile the automatic activity begins,
- identification of the operator liable for the system.
This is not about public disclosure of user data.
The aim would be to enable the competent authorities to analyse the scale of automation and the detection of related campaigns more effectively.
EU Transparency API – common reporting format
Another request is to make a single strategy for reporting information by the largest platforms.
Today, data on identified hazards can be presented in different ways, making it hard to compare them and automatically analyse them.
Therefore, the Senator proposes to consider the creation of EU Transparency API – a standardised interface for data transmission in a uniform format.
The strategy could include information on, inter alia:
- number of bots detected,
- the number of non-authentic accounts deleted,
- detected CIB campaigns,
- materials marked as generated by AI,
- cases of deepfake use,
- notifications made to the competent authorities.
This would make it possible to compare the situation on different platforms much faster and identify campaigns that are of a cross-border nature.
Exchange of information on threats
Impact operations do not end on 1 platform.
The same run can be conducted simultaneously on social media, information services, communicators and another digital services.
The Senator so proposes to make a European mechanics for exchanging information on detected manipulation campaigns.
One solution could be to usage existing standards utilized in cybersecurity, specified as STIX 2.1 and TAXIA. This would let existing experience and infrastructure to be utilized alternatively of building the full strategy from scratch.
This information could include, inter alia, the method characteristics of botnets, software signatures, publication patterns, run identifiers or information on the infrastructure used.
How can we identify whether the material was created by man or AI?
Another problem is the increasingly hard discrimination between man-made materials and those generated or importantly modified by artificial intelligence.
This includes texts, photographs, voice recordings and video material.
Deepfake’i – materials that can represent real people, places or events in an artificially generated or manipulated way – is of peculiar importance here.
The Senator calls for increased importance of open standards for confirming the origin of digital materials, specified as C2PA.
This technology allows you to add to digital materials information about their origin, editing past and the usage of AI tools.
So it is not a ban on publishing content generated by artificial intelligence.
The aim is that the user should be able to find out from which origin the material originates and whether it has been generated or substantially modified utilizing AI.
Audits of advice algorithms
An crucial component of the proposal is besides to draw attention to urge algorithms.
They decide to a large degree what content the user sees and in what order.
If a network of automated accounts is able to artificially make a immense number of reactions, comments or releases, it may effort to influence the way the advice strategy evaluates the popularity of a given material.
That's why the senator is proposing periodic independent investigating of algorithms' opposition to manipulation.
Such tests could include simulations of organised bot campaigns, investigating of opposition to artificially generated engagement and analysis of susceptibility to mass usage of AI content.
European Centre for Impact Operations Analysis
Another proposal is to consider setting up a specialised European analytical centre.
It would work with the European Commission, ENISA, national digital service coordinators and incidental consequence teams.
Its task could be to analyse cross-border manipulation campaigns, make bot detection methods, prepare mention models, conduct investigation and make analytical tools.
Such an institution would not replace national authorities.
Its function would primarily be to increase cooperation and exchange of cognition between associate States.
Senator emphasizes: it is not censorship
An crucial component of the full proposal is the clear emphasis that the aim of the proposed changes there is no regulation of freedom of expression or the introduction of universal monitoring of public communication.
The approach proposed in the Senator's speech focuses primarily on the method and behavioral features of the account network, not on the views presented by their users.
It's an crucial distinction.
The goal should not be to automatically delete content simply due to the fact that they were created utilizing AI or are popular on the Internet.
It is about identifying situations where an organized, automated network tries to artificially influence the functioning of the platform.
At the same time, any fresh regulation should take into account the protection of individual data, the right to privacy, freedom of expression, innovation and the competitiveness of the European digital economy.
What does the Senator propose to the European Commission?
In the final part of the speech, the Senator addresses the European Commission with 4 essential demands.
First – conduct a comprehensive assessment of the effectiveness of the current DSA and AI Act to counter automated impact operations.
Secondly – preparation of method guidelines for suppliers of very large online platforms, for the detection, classification and reporting of automated activity.
Third – Initiation of consultations with associate States, researchers, cybersecurity and digital service providers to make common European standards.
Fourthly – consideration of the preparation of a legislative initiative, which would complement existing rules with interoperable detection, reporting and analysis requirements for coordinated impact operations utilizing artificial intelligence.
Europe needs common standards
According to the Senator, the future of the European digital marketplace cannot be based solely on rules defining the work of platforms.
Common method standards are besides needed.
Today, the net knows no borders, and the run conducted by an automated network can besides affect users of many countries.
If each country and each platform uses different solutions, it will be much more hard to detect specified actions.
However, common standards could let for faster exchange of information, comparison of data and more effective consequence to threats.
This is not about creating another net control system. It is about creating mechanisms to separate between normal, legal user activity and organised, automated operations to manipulate the information environment.
Playing for the safety of digital democracy
The improvement of artificial intelligence makes the boundary between real and synthetic information increasingly hard to notice.
At the same time, technology itself is not a problem.
The problem may be how to usage it.
AI can support doctors, scientists, entrepreneurs, officials, teachers and citizens. It can besides be utilized to automate activities that previously required the engagement of large teams of people.
This is why, as the Senator points out, European digital policy must make with technology.
Digital Services Act has created an crucial basis for the work of the largest platforms for systemic risks. However, the scale and pace of the improvement of generic artificial intelligence make it essential to continuously check whether existing mechanisms are inactive sufficient.
Senator Adam Szejnfeld's intervention in the European Commission is so an appeal to launch another phase of work on the safety of the European information space – 1 that combines technological improvement with the protection of democracy, civilian rights and assurance in information.
Because in a planet where artificial intelligence can make thousands of realistic content in seconds, digital safety will increasingly depend not only on this, what has been publishedBut besides from this, who, how, on what scale and utilizing which tools has led to the dissemination of a given message.
Maximilian Rice
Advisor to Senator Adam Szejnfeld
