The role of the OECD in evaluating the impact of artificial intelligence on the future of work
- Author(s):
- Parliamentary Assembly
- Origin
- Assembly
debate on 1 October 2026 (35th sitting) (see Doc. 16472, report
of the Committee on Political Affairs and Democracy, rapporteur:
Ms Marietta Karamanli; and Doc.
16492, opinion of the Committee on Social Affairs, Health
and Sustainable Development, rapporteur: Ms Maria-Gabriela Horga;
and Doc. 16489,
opinion of the Committee on Migration, International Protection
and Economic Co-operation, rapporteur: Ms Larysa Bilozir). Text
adopted by the Assembly on 1 October 2026 (35th sitting).Provisional
version subject to editorial review
1. The rise
of artificial intelligence (AI) and related digital technologies
is ushering in a new era offering significant opportunities but
also posing major threats. Societies and economies must be equipped
for the ongoing AI and digital transition, in order to keep up with
the pace of technological advancements, harness their benefits and
mitigate their risks. The Parliamentary Assembly of the Council
of Europe, enlarged to include delegations from the national parliaments
of the Organisation for Economic Co-operation and Development (OECD)
member States which are not members of the Council of Europe, as
well as a delegation from the European Parliament, believes that
this requires urgent and continuous attention due to its significant impact,
notably on people, the economy, labour markets, education, public
services and governance.
2. AI is already transforming the ways in which people interact,
study, conduct research, organise production, and work. It has the
potential to become a general-purpose technology with far-reaching
effects across a wide range of sectors. Consequently, control over
AI technology and infrastructure, as well as the large data sets
required for its operation, has major strategic and geopolitical
implications.
3. To address these challenges, both the OECD and the Council
of Europe have adopted a range of instruments. In particular, in
September 2024, the Council of Europe Framework Convention on Artificial Intelligence
and Human Rights, Democracy and the Rule of Law (CETS No. 225),
the first international treaty of its kind in this field, was opened
for signature. The Council of the OECD adopted in 2019 and amended
in 2024 a Recommendation on Artificial Intelligence, containing
the following Principles for responsible stewardship of trustworthy
AI: inclusive growth, sustainable development and well-being; respect
for the rule of law, human rights and democratic values, including
fairness and privacy; transparency and explainability; robustness,
security and safety; and accountability.
4. The enlarged Assembly held its last debate in January 2024,
with the adoption of
Resolution
2526 (2024) on “Globalisation in times of crisis and war: the role
of the OECD since the Russian Federation’s aggression against Ukraine”,
which called on the Council of Europe and the OECD to continue working
together in the field of AI. The enlarged Assembly emphasises that
the Council of Europe, the OECD and their member States must be
prepared to harness the benefits of AI and related digital technologies,
while anticipating and countering any disruption to human rights,
democracy, the rule of law, and the environment.
5. The enlarged Assembly welcomes the OECD’s ongoing work to
develop a wide range of tools, repositories, indicators and research
programmes on AI. These contribute to the monitoring of policies,
the measurement of impacts, the identification of incidents and
hazards, the assessment of capabilities, and the provision of information
to governments, parliaments, businesses, social partners and citizens.
6. According to the OECD, while the potential impact of AI on
productivity growth is expected to be positive, it remains difficult
to predict its overall magnitude. AI’s impact is likely to be uneven
across economic activities: it may generate efficiency gains for
specific tasks, particularly in sectors such as finance, information
and communication technology services, media and professional services.
In contrast, sectors such as agriculture, construction, healthcare
and elder care and other activities involving a high physical and
manual intensive component, may be less exposed to AI. The extent
of these productivity gains will depend on the rate at which AI
is adopted, a process that can be accelerated by strengthening the
relevant infrastructure, data and skills. It will also depend on
the availability of complementary technologies, and on the capacity
of firms and public authorities to integrate AI responsibly and
support sectoral transformation.
7. Over the next 10 years, significant differences in macroeconomic
gains related to AI are expected among OECD countries. The enlarged
Assembly is concerned that the benefits of AI could be concentrated
among countries, regions and firms that already have stronger digital
infrastructure, higher-skilled workers, and greater financial and
institutional capacity. Differences in rates of adoption of AI between
capital regions and large cities, rural areas and remote, peripheral
or underdeveloped regions, innovation leaders and laggards, and between
large firms and small and medium-sized enterprises (SMEs), are already
apparent and may exacerbate existing inequalities. Europe’s weaker
position compared to the United States and China in terms of AI
investment, infrastructure, computing capacity and advanced models
raises questions regarding economic growth, democratic governance
and oversight, the safety of both personal and commercial data,
and digital sovereignty.
8. The precise impact of AI on employment also remains difficult
to assess and is likely to be mixed. AI is expected to create new
jobs, while simultaneously destroying existing ones, reconfiguring
occupations and changing the organisation of work. According to
the OECD, occupations at the highest risk of automation account
for an average of 27% of employment in OECD countries, though some
regions display much higher percentages. At the same time, employment
has grown in the occupations most exposed to AI.
9. According to OECD evidence, AI can improve job quality by
improving performance, reducing physical strain, automating routine
tasks, and promoting inclusion, for example for people with disabilities.
However, excessive algorithmic management, intrusive monitoring
of workers, and opaque decision making, combined with insufficient
protection of workers’ physical and mental health, could have negative
repercussions on job quality and workers’ rights.
10. The enlarged Assembly emphasises that technology alone does
not determine whether jobs will be lost or created by AI. To achieve
a favourable outcome, a combination of responsible decisions by
companies, social dialogue with workers, as well as labour reskilling,
upskilling and reallocation policies will be needed.
11. Education and skills training policies are central to ensuring
a fair digital transition. The enlarged Assembly welcomes the development
by the OECD and the European Commission of the AI Literacy Framework
for Primary and Secondary Education, which aims to equip students
with the knowledge, skills, and attitudes necessary to understand
and use AI safely and effectively. AI, particularly generative AI
(GenAI) tools, can enhance performance, learning processes, teaching
and tutoring. At the same time, these tools may also lead to overreliance,
hindering the development of independent and critical thinking.
12. The enlarged Assembly is concerned that the adoption of AI
tools in the workforce may exacerbate existing inequalities among
workers. Women are underrepresented in occupations with the highest
exposure to AI and, at the same time, they are overrepresented in
certain occupations that may be at a higher risk of automation,
such as clerical roles. Furthermore, women are less likely to participate
in science, technology, engineering and mathematics (STEM) fields.
Inequalities may also emerge among different age groups: young people
are particularly vulnerable, as entry-level jobs are more easily
automated, while older people are at risk as they are less likely
to adapt to using AI tools. Migrants, refugees and asylum seekers
face comparable risks of exclusion and discrimination, in particular
where AI-based tools are used without adequate safeguards, for recruitment,
identity verification, skills-matching or residence and work-permit
decisions affecting them. AI technologies, however, have the potential
to significantly improve the inclusion of people with disabilities
in the workforce.
13. Cultural and creative labour employment are among the most
exposed to GenAI. High exposure does not necessarily mean job destruction,
but it does require labour market policies to support shifting skills demands.
OECD work on AI highlights the importance of policies that support
lifelong learning, responsible AI adoption that complements human
labour, protection of job quality and the broad sharing of productivity
gains. This is especially important in cultural and creative industries,
where self-employment, project-based work, micro, small and medium-sized
enterprises, precarious conditions and limited social protection
can make adapting to technological change more difficult.
14. Other reasons for concern are cybersecurity risks, and potential
issues with copyright and intellectual property stemming from the
use of GenAI tools. This is particularly concerning given the use
of “data scraping” practices, which involve the automatic extraction
of public or non-public data and information from third-party sources
to train AI models. The risks from these practices could be sharper
for SMEs, particularly in creative industries, which have more limited
resources to protect their data. Fair compensation for the use of copyrighted
material in training models, clearer licensing frameworks, and stronger
transparency around the use of creative works in AI systems will
therefore be important not only for equity, but also for maintaining
the long-term sustainability of creative careers and cultural production.
15. The media and journalism are also affected, as the uncontrolled
use of AI can have repercussions on the quality of information,
on pluralism, and, ultimately, on democracy. In intellectual, creative,
and journalistic professions, the goal should not be to produce
more content at a lower cost, but to use AI to free up time for investigation,
creation, interpretation, fact-checking, human interaction, and
editorial responsibility
16. The OECD is also conducting research into the use of AI by
public authorities in core government functions. The enlarged Assembly
recognises that the full potential of AI tools in different public
services remains untapped, and that careful design, implementation
and monitoring will be required to mitigate risks.
17. The enlarged Assembly believes that Europe needs to innovate
and strengthen its digital ecosystem to reduce the gap with countries
such as the United States and China in terms of digital infrastructure
(data centres; hardware; computing capacity; advanced models) and
investment, and to prevent growing disparities across firms, sectors
and places. This means fostering open and fair trade, which facilitates
access to technology without increasing Europe’s dependencies or
reinforcing the dominant positions of a few large companies. Europe
must therefore promote effective competition, data interoperability,
fair access to essential digital infrastructure, and common rules
to regulate AI and monitor its development. The enlarged Assembly stresses
the necessity to invest in order to ensure European AI sovereignty.
Given that AI technologies are avid energy and water consumers,
Europe must also ensure that the digital transition is balanced
by environmental considerations.
18. In light of the above, the enlarged Assembly invites member
States of the Council of Europe and the eligible member States of
the OECD to sign and ratify the Framework Convention on Artificial
Intelligence and Human Rights, Democracy and the Rule of Law in
order to allow its rapid entry into force, and encourages non-member
States worldwide to request to accede to it, as soon as they have
the opportunity to do so.
19. The enlarged Assembly further invites member States of both
the Council of Europe and the OECD to:
19.1 adopt and implement coherent national strategies and regulatory
frameworks for AI and digital transformation that are human-centred,
rights-based, socially inclusive and environmentally sustainable, in
line with the Council of Europe Framework Convention on Artificial
Intelligence and Human Rights, Democracy and the Rule of Law and
with the OECD Principles for responsible stewardship of trustworthy AI;
19.2 ensure that the Council of Europe Framework Convention
on Artificial Intelligence and Human Rights, Democracy and the Rule
of Law applies to all business entities operating on their territory
in a manner consistent with the requirements of the European Social
Charter (ETS No. 35 and ETS No. 163), the European Union AI Act,
the OECD Principles for responsible stewardship of trustworthy AI
and the related Due Diligence Guidance for Responsible AI, relevant
International Labour Organization (ILO), conventions and national
labour law;
19.3 establish monitoring mechanisms to assess the impact of
AI on productivity, the labour market, wages, skills, job quality,
occupational safety and health, equality, intellectual property
and regional development;
19.4 sustain broad-based and open-source responsible AI diffusion
and foster its effective integration, which will be key to maximising
labour productivity and workers’ income in the medium term. Key
levers for achieving this would include investing in computing,
connectivity and complementary skills, and having in place broader
economic framework conditions that allow firms and workers to adopt
and experiment with the technology;
19.5 explore the feasibility of an international treaty that
would prohibit the development of so-called “artificial superintelligence”
technology and other advanced AI technologies worldwide until appropriate guardrails,
regulatory frameworks and a multilateral supervisory body are in
place to guarantee their safe use and full human control.
20. Regarding the impact of AI on the labour markets, the enlarged
Assembly invites member States of both the Council of Europe and
the OECD to:
20.1 support investments
in human capital to help workers, jobseekers and employers seize
the benefits of AI, navigate changing skill needs and demands and
transition to new jobs;
20.2 encourage the development and deployment of AI systems
designed to complement human capabilities rather than diminish them,
particularly by delegating repetitive or low-value-added tasks, supporting
the detection of anomalies, improving workplace accessibility, and
allowing workers the opportunity to understand, question, correct,
or disregard the recommendations generated by the system;
20.3 support efficient reallocation of labour across occupations
and industries, including SMEs, as sluggish reallocation significantly
limits the productivity benefits of AI. This involves facilitating
the transition of workers affected by AI-driven productivity changes,
while sustaining activity in less exposed sectors. This will require
secure labour-market transition arrangements, strong social protection, targeted
retraining and mobility support, developed through social dialogue
and in consultation with workers and their representatives;
20.4 promote awareness and building knowledge of the opportunities,
functionalities, potential uses, limitations and impacts of AI systems
in the labour market among workers (including managers) and employers,
notably in SMEs, including possible biases and negative impacts
on job prospects;
20.5 conduct an impact assessment involving workers and their
representatives, prior to the deployment of AI systems that are
likely to have significant effects on workers or job seekers. This assessment
should, in particular, examine the effects of the AI system on the
organisation and intensity of work, autonomy, skills, compensation,
health and safety, the protection of personal data, the risks of discrimination,
the procedures for challenging decisions, and the allocation of
responsibilities among the employer, the provider, and the user
of the system;
20.6 promote social dialogue through the appropriate and effective
engagement of workers, trade unions, employers and their collective
organisations, including through collective bargaining, to facilitate the
trustworthy use of AI in the labour market;
20.7 ensure the protection and respect of labour rights, including
through regulation where necessary, and support employers in taking
effective measures against potential labour rights abuses at all
stages of the AI system lifecycle in the labour market, while involving
workers and their representatives in the design, implementation
and evaluation of such measures;
20.8 consider developing appropriate measures to preserve workers’
rights to freedom of association, collective bargaining and privacy
where the integration of AI systems in the labour market entails
the collection of personal data from workers and jobseekers, in
a proportionate manner, based on evidence of the benefits and risks;
20.9 consider opportunities to leverage the use of AI to monitor
and mitigate bias and discrimination in the labour market, to ground
decisions on disaggregated data and evidence, and to help ensure
that AI systems produce results that do not perpetuate and exacerbate,
as a magnifying glass, entrenched systemic inequalities;
20.10 support workers and employers in using trustworthy, safe
and secure AI to prevent work-related injuries and ill health, and
to improve job quality, while also addressing the risks for occupational
safety and health in a proportionate manner;
20.11 promote transparency, explainability, accountability and
democratic approaches in the deployment and use of AI systems in
the labour market, including in SMEs, taking a proportionate approach
based on evidence of the benefits and risks, to help workers and
employers understand how the systems work, what the outputs are,
which are the cybersecurity risks, and who is responsible for ensuring
they function properly;
20.12 ensure that no decision regarding hiring, assignment,
evaluation, compensation, promotion, disciplinary action, or termination
of employment is made solely on the basis of an automated system without
qualified human intervention, and guarantee that when automated
systems are involved, this is comprehensibly explained to affected
individuals and effective legal remedies are provided, if needed;
20.13 seek to address gaps in territorial and regional infrastructure,
data, skills and capacities for AI adoption, paying particular attention
to the different challenges facing local and regional labour markets to
ensure all regions have the potential to benefit from new tools;
20.14 strengthen local labour market policies and local governments’
capacity to understand the impact of AI on work, to support workers’
adaptation, and to promote training;
20.15 promote through social dialogue and collective bargaining,
the fair sharing of AI-driven productivity gains and explore and
encourage the responsible use of AI to reduce the amount of employee’s
working hours, while maintaining productivity and growth, in order
to support their work-life balance and improve their social well-being;
20.16 encourage voluntary and market-based mechanisms that enable
employees to participate in the economic opportunities created by
AI-driven productivity gains, including employee share ownership, share-option
schemes and other forms of long-term financial participation, while
fully respecting entrepreneurial risk, property rights and the freedom
of companies to determine their ownership and remuneration structures;
20.17 ensure that AI-based tools used in recruitment, employment
and residence- or work-permit procedures affecting migrant workers
are subject to meaningful human oversight, are regularly audited for
discriminatory bias, and comply with applicable international human
rights and data protection standards;
20.18 ensure that AI-driven labour-matching and job-allocation
tools are designed and monitored in a manner that avoids reinforcing
the occupational segregation of migrant, refugee and other foreign workers
into low-quality or precarious jobs, and actively supports their
fair access to skilled employment matching their qualifications.
21. With regard to the impact of AI on the creative industries,
the media, and journalism, the enlarged Assembly calls on the member
States of the Council of Europe and the OECD to:
21.1 develop policies aimed at protecting
businesses in the creative industries, in particular by revising intellectual
property legislation and by regulating practices such as data scraping.
The measures to be taken should also address the unauthorised use
of protected content, the production of deepfakes, the standardisation
of cultural products and the possibility of human creators being
replaced by AI. The aim should be to limit the use of AI tools that
would lead to the standardisation of language, formats, references,
narratives and tastes, to the detriment of Europe’s cultural, linguistic
and intellectual diversity;
21.2 protect public-interest journalism, media pluralism and
the integrity of information by ensuring that any use of AI systems
in newsrooms complies with the principles of truthfulness, accuracy, independence,
editorial accountability, protection of sources and pluralism;
21.3 ensure that no content generated or substantially modified
by AI is published without effective human verification and without
a clearly identified natural or legal person assuming editorial responsibility
for it;
21.4 to invite media organisations and bodies representing
journalists to negotiate collective agreements governing the introduction
of AI, covering in particular the provision of information to and consultation
with employees, permitted and prohibited uses, the protection of
sources and confidential data, verification procedures, the labelling
of synthetic content, the safeguarding of jobs and training pathways,
as well as fair remuneration for the use of journalistic content.
22. Regarding the strengthening of education and training capabilities,
the enlarged Assembly invites member States of both the Council
of Europe and the OECD to:
22.1 invest
substantially in and promote the development of lifelong learning
on AI literacy to equip students and workers with the knowledge,
critical thinking skills and ethical awareness needed to understand
and use AI effectively and responsibly;
22.2 design policies to ensure that women and men have equal
access to AI tools and AI-related STEM education and training. Special
attention should be paid to low-skilled workers, young people in entry-level
jobs, older workers, persons with disabilities, and regions with
more vulnerable employment structures;
22.3 ensure that the use of GenAI in education prioritises
long-term learning over short term performance gains, including
by using general-purpose GenAI with pedagogical intent where appropriate,
co-designing specialised educational AI tools involving teachers
and students, and providing guidance to prevent overreliance, which
could hinder cognitive development;
22.4 support teachers and education systems in leveraging AI
as a complementary and augmenting tool, while safeguarding the human
dimension of teaching, strengthening professional capacity and fostering
evidence-based policies for the use of AI in classrooms;
22.5 promote dialogue and co-operation between local vocational
education and training providers, innovation hubs and universities,
industry and business representatives (including SMEs), workers’ representatives
and policy makers, to better anticipate AI-driven skills demand,
provide reskilling and upskilling opportunities, and align future
supply capacities.
23. Recognising the potential impact of AI on the way the public
sector operates and delivers services to citizens, including national
and subnational governments, the enlarged Assembly calls on member
States of both the Council of Europe and the OECD to:
23.1 support public administrations
at all levels of government in preparing for AI through strategic
civil service workforce and activities planning. This should include
defining which tasks can be delegated to AI, assessing the impact
on job roles and staffing, identifying future skills requirements,
and investing in the reskilling and upskilling of civil servants
and human resources professionals, as well as the recruitment of
AI-literate individuals;
23.2 ensure that the use of AI in public-sector human resource
management, including recruitment, promotion, learning and performance-related
processes, remains merit-based, be transparent, explainable and
contestable, with qualified human oversight and accountability for
final decisions;
23.3 protect the job quality, autonomy, privacy and professional
discretion of civil servants at all levels of government where AI
is used in management, workflow allocation, monitoring or productivity
tools, including by outlawing intrusive surveillance, and involving
employees and their representatives in the design, implementation
and evaluation of such systems;
23.4 promote the use of AI to reduce administrative burdens
and strengthen learning and development in the public sector, while
monitoring and evaluating its impact on fairness, productivity,
employee well-being and service quality;
23.5 strengthen AI-related capabilities among senior leaders
and decision-makers in the civil service, to ensure that they have
a sufficiently detailed understanding of how AI systems function,
their limitations, risks and operational implications, with a view
to enabling them to make responsible and informed decisions on the
use of AI in the civil service;
23.6 address the broader transformation of work driven by AI
adoption in the public sector, including changes to tasks performed
by humans, the redesign of roles and workflows, and the emergence
of new job roles, and assess the impact on working conditions, job
quality and the operating environment;
23.7 expand public-sector learning and development systems
beyond basic AI literacy and ethics, which are well established
now in many governments, to support the integration of AI into longer-term government
operations. This should include the development of technical and
role-specific skills, procurement knowledge, and substantial reskilling
and upskilling strategies. Overall, the capacity of schools of government
should be strengthened to enable them to autonomously design and
deliver practical AI training;
23.8 provide targeted support for subnational governments by
promoting data access and sharing across jurisdictions, facilitating
interoperability, and strengthening digital infrastructure and workforce capabilities.
24. The enlarged Assembly invites the OECD to continue and deepen
its work on measuring and evaluating the economic, social, territorial
and environmental impacts of AI. The OECD should continue to make
its evidence accessible to parliaments, through initiatives such
as the OECD Parliamentary Group on Artificial Intelligence, and
should provide support to policy makers in defining practical, rights-compatible
and socially inclusive reforms. The OECD should also strengthen
its co-operation with the Council of Europe in the field of AI,
in particular by promoting complementarity between Council of Europe
legal standards and OECD policy tools.
25. The enlarged Assembly calls on the OECD to develop specific
indicators for measuring the subnational, national and regional
dimension of the artificial intelligence transition, including disparities
in access to digital infrastructure, AI-related investment, skills
and training, the capacity of SMEs to adopt AI technologies, and
the creation of new employment opportunities, with a view to identifying
and reducing the emerging ‘regional AI divide’. The OECD is also
invited to give due consideration to the specific impact of AI on
migrant, refugee and cross-border workers.
26. Furthermore, the enlarged Assembly encourages the OECD to
deepen its analysis of the various impacts of AI on work, drawing
a distinction between, on the one hand, AI systems and tools that
assist workers, alleviate arduous tasks, facilitate problem solving
and ensure human oversight, and, on the other hand, those that dictate,
monitor and accelerate the pace of work, and replace professional
judgement with automated recommendations. In particular, the OECD
could, in co-operation with the Council of Europe, the International
Labour Organisation, the social partners, occupational health organisations
and research institutes, develop a common framework for assessing
the impact of AI on work, distinguishing in particular between “supportive”
and “alienating” uses, and enabling the measurement of their effects
on autonomy, workload, intensity, job quality, skills retention,
health, discrimination and social dialogue.
27. Additionally, the OECD is encouraged to develop a common scorecard
to compare, on a country-by-country basis, not only the uptake of
AI, but also the proportion of organisations (businesses and public authorities)
that have involved workers in its roll-out; the proportion of systems
subject to an impact assessment on rights and working conditions;
the effects on wages, working hours, work rates, health and autonomy;
and actual access to training, broken down by gender, age, qualifications,
employment status, region and company size.
