The European Economic and Social Committee (EESC) has adopted an opinion that aims to safeguard citizens’ fundamental rights in the midst of the implementation of AI technology in public services.
Artificial Intelligence (AI) is becoming more widespread in society and the digital transition has changed ways of working – including the organisation of work itself. These changes have a significant impact on society, the economy, production and work. Though the development of AI contains potential benefits, it nevertheless raises concerns about the effects it could have on employment levels and working conditions.
Public services are also impacted by this new technology and there are already several cases of public services in the EU using AI. Though AI offers (among others) the possibility to automate complex and repetitive processes and help make public services more accessible to citizens, generative and predictive AI algorithms may (if unchecked) lead to bias.
The EESC’s approach
The EESC has adopted an opinion focusing on the use of AI in public services, providing a series of recommendations on how best to implement and integrate AI in this crucial sector, noting that this technology could “revolutionise the way public services are provided, offering innovative solutions for dealing more efficiently and more quickly with people’s needs”.
The opinion recommends upholding transparency in the decision-making processes surrounding the creation of algorithms and the importance of guaranteeing the human-in-command principle, keeping in mind that public services are provided by human beings, at their own pace and with their own requirements.
Public service employers should also ensure their workers are aware of the adoption of AI monitoring systems – especially for those working directly with these systems. This will help ensure there is transparency surrounding administrative activities involving AI: clear information surrounding the implementation of such systems is key to fostering trust regarding this new technology and its use.
A European way of using AI
In January 2024, the European Commission defined a framework for the AI systems it is using and published a Regulation (AI Act) laying down harmonised rules on artificial intelligence. Meanwhile, the EESC
recommended transparency in decision-making processes when developing algorithms – even before the adoption of the AI Act.
These legislative steps demonstrate that a European approach to using AI and digitalisation is emerging, namely one that strikes a balance between technological development on the one hand and safeguarding citizens’ rights on the other. It is an approach that adopts a human-centric point of view and promotes dialogue with relevant civil society stakeholders (consumers’ and patients’ rights organisations, minorities’ rights representatives, etc.).
The EESC opinion underlines the fact that, in order to develop a secure AI system that respects fundamental rights, a climate of trust must be fostered around the use of this new form of technology. Social dialogue and collective bargaining should play a crucial role in creating this climate and innovations adopted should be regularly monitored and supported by social partners.
Boosting skills and security
A recent OECD study has illustrated that, unlike previous technological advances, AI presents completely new features, meaning upgrading the skills of citizens and workers is crucial for society to successfully integrate this technological transition. Though the reliance on some skills based on more repetitive tasks may become less needed over time, new skills linked to developing and maintaining AI systems will become more important. The introduction of AI in public services will therefore have to be coupled with the creation of comprehensive training and upskilling plans.
Public services manage a huge amount of data which may include sensitive information (such as individuals’ health and legal data). In the EESC’s view, the introduction of AI in this sector should therefore also be accompanied by reinforced levels of security in data acquisition, processing and storing. On top of this, public services should ensure they have the cybersecurity tools in place to prevent attacks and other data crimes.
EESC’s recommendations and conclusions in full
1. Conclusions and recommendations
1.1 ‘Public services’, as such, are not defined in the Treaties, which refer instead to Services of general (economic) interest (SGIs) in Article 14 TFEU and Protocol 26 TFEU. In this opinion ‘public services’ is taken to mean the activities or services that public authorities of the Member States at national, regional or local level classify as being of general interest and, therefore, subject to specific public service obligations (PSO).
1.2 Public services are fully involved in the digital transition, and in order to make the best use of new technology, it is necessary to keep in mind its advantages and disadvantages.
1.3 Adopting digitalisation and AI (artificial intelligence) could revolutionise the way public services are provided throughout the European Union, offering innovative solutions for dealing more efficiently and more quickly with people’s needs.
1.4 AI can improve the provision of services to users and help make these services more accessible, in particular to the most vulnerable people in society.
1.5 The possibility to automate complex and repetitive processes could increase efficiency while reducing workload for individual workers. However, care must be taken with the use of predictive and generative artificial intelligence algorithms – for example, to prevent bias – so that the systems used are fair and transparent, and safeguard the rights of all actors involved.
1.6 The EESC recommends transparency in decision-making processes when developing algorithms1 and in the implementation of these technologies, also to guarantee the human-in-command principle. While improving the efficiency of public services is important, it should be borne in mind that these services are provided by human beings, at their own pace, in their own time and with their own requirements.
1.7 Public services must invest in and ensure there are the cybersecurity tools needed to prevent attacks and other data crimes.
1.8 Public service employers must make their workers aware of the adoption of AI monitoring systems in order to ensure the transparency of administrative activities both internally and externally, especially for those working with the new systems. Information is crucial to creating trust, as well as for training. The integration of AI into public services should therefore be accompanied by a comprehensive training and upskilling plan.
1.9 Social dialogue and collective bargaining play a crucial role in fostering the transition to artificial intelligence. The innovations adopted need to be accompanied, directed and conveyed through regular monitoring by and support from the social partners.
1.10 To ensure that the digital transition and the introduction of AI in public services is inclusive and sustainable, it is necessary that there be dialogue with relevant civil society stakeholders such as consumer rights organisations, patients’ and disability organisations, and minority rights representatives.
2. General comments
2.1 We are in the midst of complex societal challenges, several which have been fuelled by events such as the recent global pandemic and multiple crises. The digital transition in particular has changed ways of working and the organisation of work itself. Artificial Intelligence (AI) is becoming more widespread in society, with a significant impact on society, the economy, production and work. While the scale of the changes taking place highlights the opportunities that AI can provide in terms of improving economic and social systems, it also raises doubts and concerns regarding the protection of fundamental and social rights.
2.2 Some early studies on the use of AI in the workplace2 have illustrated that, unlike technological changes in the past, AI presents completely new features. AI systems have a high degree of autonomy and operate with different levels of automation. Whilst certain generative AI tools can potentially operate without human intervention, the majority of AI tools do not use fully‑automated systems.
2.3 Some of the main motivations for employers to adopt AI include its potential to increase efficiency and productivity and improve the quality of products and services provided. Workers can also benefit from the introduction of AI. It can improve the organisation of workflow, eliminate dangerous or repetitive tasks and create more complex and interesting ones, thereby increasing workers’ autonomy. As well as benefits, however, the development of AI also raises concerns about the effects it could have on employment levels and working conditions.
2.4 Public services are also fully involved in these momentous changes and, like the rest of the working world, are being called upon to make the best use of this new technology, while keeping in mind its advantages and disadvantages. There are already several cases of public services in the EU using AI, assessing its benefits and drawbacks. The European Commission has itself recently defined a framework for the AI systems currently in use at the Commission, as well as for those in the process of being developed, and action has been taken to ensure that AI is lawful, safe and trustworthy3. In the context of public services, AI can improve the provision of services to customers and can help make those services more accessible, in particular for the most vulnerable people in society.
2.5 Adopting digitalisation and AI could revolutionise the way public services are provided, offering innovative solutions for dealing more efficiently and more quickly with people’s needs. The use of certain forms of AI, such as machine learning, has made it possible to process large data volumes and thereby increase the number of cases processed, reducing waiting times for services provided to citizens. For activities that do not require contact with the public, remote working has been introduced, allowing public services employees to organise their own work, including in relation to their own personal needs.
2.6 In this context, the newly adopted AI Regulation provides, for the first time, a common legal basis for EU countries to guide the process of introducing AI, in order to make it socially sustainable so that it does not infringe fundamental rights. The new rules set the conditions and limits for the use of artificial intelligence based on risk levels, reaffirming the human-centric vision of AI. We need to continue down this path and promote European digital sovereignty, as technological autonomy can help ensure compliance with European standards, such as data protection, inclusion and respect for rights.
2.7 This EESC exploratory opinion requested by the European Commission aims to shed light on the impact of digitalisation and on the introduction of artificial intelligence in public services. Since public services are not defined in the Treaties, this opinion is targeted only towards activities provided by public entities, which are referred to as ‘public services’ in this document. This opinion assesses how AI and digitalisation can be used as a tool for improving the efficiency of public services and the organisation of public work, by examining how AI and the use of new technologies could change the relationship between users and public services, while taking into account work changes that could affect workers in public services. Finally, this opinion will also look at how AI can contribute to more equal and inclusive societies.
3. The impact of AI on public services
3.1 The application of AI in public services deserves particular attention, especially with regards to elements pertaining to the protection of citizens’ rights and the management of public goods and services, which are the focus of this opinion. This requires a particular focus on the conditions under which AI is adopted and used, to make it compatible with the mission, nature and objectives of public action. The transition to a digital world, for example, must not entail bypassing standards.
3.2 The wide variety of areas of activities is matched with a broad coverage of ‘public tasks’ which concern decision-making processes, resource allocation, the definition of measures, the actual provision of services and the organisation of work. Therefore, when considering the effects of AI use, specific focus should be placed on the sector in question and the type of task concerned. For example, consider the complex structuring of public services provided by public administrations. These can involve activities of a strictly administrative nature conducted by central governments and local and regional authorities, but also public education, the provision of healthcare services, defence, social security, justice and tax collection.
3.3 The extent to which public services are highly conditioned by national laws should also be taken into consideration. Such laws establish objectives, methods of operation and rules of conduct, which vary not just between countries but also sometimes within individual countries themselves. This makes it more difficult, at European level, to identify indicators that can be applied to every context and region, making it particularly challenging to establish coordinated governance on AI issues.
3.4 The possibility to automate complex and repetitive processes could increase efficiency while reducing workload for individual workers. However, care must be taken with the use of predictive and generative artificial intelligence algorithms – for example, to prevent bias through independent bias audits – so that the systems used are fair and transparent, and safeguard the rights of all actors involved. As highlighted by the Joint Research Centre4, public services can benefit greatly from the use of AI, improving decision-making processes and policy outcomes, enhancing public service delivery and user/government interaction, and optimising internal management. For example, at local level, the use of AI is increasing in smart cities due to the enormous amounts of data generated by sensors, the Internet of Things (IoT), citizens, etc. Many cities are creating their own AI-enabled local digital twins to support the goals of the Green Deal, focusing on a single domain (e.g. the LEAD project on urban logistics.
3.5 The EESC had already recommended transparency in decision-making processes when developing algorithms, before the adoption of the AI Act. In order to guarantee the human-in-command principle, the forthcoming ethical guidelines on AI to be prepared by the Commission should provide for the interaction between workers and intelligent machines and establish principles of participation, responsibility and ownership of production processes. While improving the efficiency of public services is important, it should be borne in mind that these services are provided by human beings, at their own pace, in their own time and with their own requirements.
4. Accountability and security (cybersecurity) in public services
4.1 Introducing forms of AI into public services entails challenges that must be addressed. This transition will require significant organisational efforts by the public sector, which is not always up to the task, given the enormous quantities of data involved, the lack of staff training and the fact that, in some cases, the IT systems concerned have not yet been fully modernised. The introduction of AI also requires that the environmental impact of such technologies be taken into consideration, given that, on the one hand, machine learning systems consume more energy, yet on the other, AI can be used to find solutions to environmental challenges5.
4.2 When introducing AI, public services must ensure that adequate information is provided to citizens, businesses, workers and other stakeholders concerned. At the same time, it is essential to clarify where responsibility and accountability lie, by identifying those responsible for managing the administrative process using artificial intelligence solutions and by ensuring a focus on potential bias in the selected solutions. This is in order to prevent AI use from depriving citizens, users, and beneficiaries of public services of a human contact person they can go to for advice or to report incorrect options generated by algorithms.
4.3 In view of the above, forms of strategic partnership between the public and private sectors are needed that create digital tools capable of providing public services more efficiently. However, public services should be accountable and retain governance and control of the process, expressing their own needs and the goals they intend to achieve, potentially relying on qualified private entities to provide these services. Special procedures (or criteria) should be established as early as the algorithm design stage to ensure that public services are fair and citizens’ rights protected, preventing incorrect, discriminatory or flawed outcomes. This should extend to the use of data in training or feeding the algorithms. Similarly, when feeding these algorithms, steps must be taken to ensure that all the spheres concerned by these decisions are fairly represented and the ethics considered. The above-mentioned transparency of decision-making processes for creating algorithms is thus becoming essential, especially where this could affect access to social and health rights.
4.4 The enormous amount of data managed by public services might include overly sensitive information (for example health and legal data or data concerning social rights). Considering the specific nature of the information and data processed, the introduction of AI must be accompanied by reinforced security levels in data acquisition, processing and storing. Public services must invest in and ensure there are the cybersecurity tools needed to prevent attacks and other data crimes. This includes monitoring unusual network activities and access points, identifying potential data vulnerabilities and reinforcing restrictions on access to sensitive data, improving the accuracy of intrusion detection systems, identifying system vulnerabilities, and regularly updating security systems. The GDPR provides personal data protection and must be fully implemented and considered throughout development.
5. AI in the world of public services
5.1 Given the impact that the introduction of AI has on work, public services will be required to take the relevant steps to ensure that this occurs in the best way possible.
5.2 Primarily, AI development in public services must be based on prior discussions with both the public entities that provide the services and the trade unions who represent the workers in these public entities. As public service provision is often gendered, with migrant women more often in the least skilled jobs, efforts should be undertaken to ensure that such workers are also represented. Together, public entities at local, regional and national levels, and trade unions, can identify any problems relating to the organisation model proposed and put forward their ideas on how to improve it and how to protect workers. In any case, the introduction of AI in public services should follow national rules on information and consultation.
5.3 The introduction and development of AI is based on the involvement of workers, following the human-centric approach mentioned above. In this context, it is essential to both recruit and train workers, for any digital innovation, as regards its potential impact on fundamental rights, in particular for AI. Regarding recruitment, it should be kept in mind that some Member States have a deficit of workers in public services, and this could significantly hinder the implementation of digital services and the provision of services. The number of public services workers should therefore be increased accordingly where their numbers are not sufficient to ensure the provision of public services. It is also necessary to include specific job profiles for workers capable of managing this delicate process (such as IT experts or engineers). With regard to training, it should be remembered that simply hiring staff is not enough; public services workers must also be trained so that they learn how to interact with the new platforms and with the new ways of working.
5.4 In addition to training, ensuring that the workers involved are well-informed is also crucial. Public service employers must make their workers aware of the adoption of AI monitoring systems to ensure the transparency of administrative activities internally as well as externally, particularly for those working with the new systems.
5.5 The use of AI could have a negative impact on workers’ health6. For example, working from home remotely can increase isolation and reduce interaction with colleagues. Other crucial aspects include an increase in psychosocial risks due to workload, as well as working times that are dictated by machines. The number of hours workers are connected will increase, and the boundaries between personal and professional life will become more blurred as workers have to guarantee efficiency7. Moreover, the fact that a machine – not another human being – will be giving instructions runs the risk of employees feeling alienated. In this respect, it would be useful for the Commission to consider how to follow up on the European Parliament resolution of 10 March 20228 calling for a directive on psychosocial risks to be adopted. The EESC is ready to issue an opinion on this matter. At national level, training for operators in the sector, including labour inspectors, is also important, so that they can monitor this risk and protect the workers involved.
6. Ensuring rights and social inclusion and promoting democratic participation
6.1 As the approval of the AI Act has already demonstrated, there is a European way to using AI and digitalisation that strikes a balance between technological development and citizens’ rights.
6.2 When applying AI solutions, user rights must be fully considered, since users are the recipients of the services. There is a risk that, in the race for efficiency, AI solutions become too general, built for general populations rather than taking specific user groups’ or individuals’ specific characteristics into consideration. Therefore dialogue with the relevant civil society stakeholders – such as consumers’ and patients’ rights organisations, and those representing the rights of women, people with disabilities, racialised people, LGBTQI+ people, and other minorities – should be promoted. Given the highly gendered nature of public service provision, coupled with gender inequalities and given the health differences between the genders, it is important to include a gender-lens that takes an intersectional approach to other specific characteristics. This will ensure inclusive and sustainable systems that are tailored to the issues and/or needs involved, since risks and opportunities can be different for different groups. Inclusion can be achieved through, for example, improved access to information, services tailored to the needs of the diverse populations, training, and support to help marginalised groups access and use AI-driven services.
6.3 From this point of view, which must only be a human-centric one9, AI development must recognise the fact that there may be sections of the population that could be excluded from these benefits. Voice assistants, for instance, can replace the textual interfaces of public services’ websites10, allowing not only easier use for existing public users, but also access for citizens who would otherwise not be able to interact with the service providers, such as citizens with visual impairment or elderly people with low levels of digital literacy. The risk of social exclusion should be reduced by providing possibilities to access public services through non-digital systems, as a fully computerised public services system cannot be envisaged. The use of tools such as chatbots can therefore be included in the list of options that people can use to contact public services, but they must not be the only ones available. Other options should always be provided, such as the possibility to make phone calls to staff, video calls or in-person appointments. A core objective must therefore be how the technology can be used for more inclusive societies.
7. Social dialogue to accompany the changes
7.1 To promote a secure AI system that respects fundamental rights such as privacy and fairness, transparency and understandable processes, a climate of trust needs to be created around the introduction of AI systems.
7.2 Social dialogue and collective bargaining play a crucial role in fostering the transition towards artificial intelligence. The innovations adopted need to be accompanied, directed and conveyed through regular monitoring by and support from the social partners. The social partners can help to achieve a better, transparent and rights-based use of AI by favouring a flexible and pragmatic approach that also promotes fairness and transparency.
7.3 Social dialogue can also play a key role, including in terms of how work is organised. From this point of view, ways of working that became firmly established during the pandemic, such as remote working, could be positive and compatible with more efficient delivery of public services. In this context the European social partners framework agreement on digitalisation, signed on 22 June 202011 by the ETUC, BUSINESSEUROPE, SME United and CEEP, and the European sectoral agreement on digitalisation for central government administrations, signed on 6 October 202212 by TUNED and EUPAE are relevant.
8. Boosting digital skills for all
8.1 Upgrading the digital skills of citizens and workers is crucial for society to successfully tackle this transition. It is therefore necessary to make citizens and all sections of civil society more aware of the opportunities and risks of AI and ensure they have greater proficiency in using the services rendered through AI systems, through vocational education and training.
8.2 As regards public services, the introduction of AI is bringing about a change in skills needs. On the one hand, it will replicate (and could partly replace) certain skills, such as the more repetitive ones that can be automated, as well as cognitive ones such as analysis, planning and advice. On the other hand, the skills needed to develop and maintain the artificial intelligence systems will become more important, as well as those needed to implement, use and interact with AI applications.
8.3 The integration of AI into public services should therefore be accompanied by a comprehensive training and upskilling plan. This should concern current job profiles in public services, where skills need to be upgraded and redefined to meet the new needs of organisations, as well as the recruitment of new workers who can provide the innovative skills needed to facilitate the transition.
8.4 Such development is contingent on massive investment of financial resources to make the entire public services of EU Member States digital. NextGenerationEU provides a unique opportunity to modernise all public services. All the resources available must therefore be used for this purpose in order to make the digital revolution a reality.
A case from the EESC