C Explanatory memorandum by Ms Maria-Gabriela
Horga, rapporteur for opinionNote
1. The rapid integration of artificial
intelligence (AI) into work environments brings not only the promise
of productivity and efficiency gains but also introduces a broad
spectrum of risks and challenges for employers, employees and society
at large. As AI becomes increasingly embedded in workplace processes,
its impact extends well beyond pure automation, reshaping job roles,
skill requirements and even the core notions of equality and well-being
at work. The report by Ms Marietta Karamanli on behalf of the Political
Affairs Committee fairly reflects most of these continued concerns
and the OECD’s contribution to developing collective responses to
ensure the responsible and safe use of AI.
2. A central risk associated with AI in the workplace is the
potential for job displacement, exacerbating entrenched inequalities.
AI systems are expected to replace existing jobs: according to the
OECD’s estimates, about 27% of occupations in member countries are
at high risk of automation, though in some regions and for certain
activity sectors this percentage is even higher. While new jobs
emerge and workflows are transformed, the uncertainty, the speed
and unevenness of this transition will leave many workers vulnerable.
Additionally, countries and enterprises already endowed with strong
digital infrastructure and high-skilled workforces are better positioned
to adopt AI, putting others at a competitive disadvantage. SMEs
(small and medium-size enterprises) and creative industries, lacking
the resources of larger firms, are more likely to struggle to protect their
workforce from these disruptive trends. (see proposed amendment
C)
3. Reckless implementation of AI in the workplace also poses
significant risks to job quality and workers’ rights. Algorithmic
management tools, which increasingly control the allocation of work,
monitor performance and even make decisions about hiring and firing,
can lead to a dehumanised workplace. The OECD has reported concerns
about intrusive worker surveillance and decision-making processes
that lack transparency or meaningful avenues for appeal. All too
often AI is used without notifying the users and without giving
them the choice of refusing AI-powered tools. As most AI applications
are developed by the business entities, short-term corporate profit-seeking
and vested interests tend to prevail over longer-term privacy, cyber
security, safety and environmental concerns.
4. Discrimination and bias are persistent risks wherever AI systems
are deployed. AI models trained on historical data may replicate
or even amplify existing prejudices, adversely affecting hiring,
performance assessments or promotional opportunities. AI-enabled
monitoring may involve continuous surveillance of employees’ communications
and physical movements, generating huge amounts of personal and
behavioural data. The boundaries between legitimate performance
management and invasive oversight become blurred, with employees
potentially subjected to permanent data collection and analysis.
Even with well-intentioned design, the complexity and opacity of
many AI tools make it difficult to ensure that they do not introduce
or perpetuate unfairness in the workplace.
5. The challenge of accountability also looms large. When decisions
with significant consequences are made by AI systems, determining
liability for errors or harm can become diffuse, involving employers,
software providers or end-users. This complicates legal recourse
for employees who feel harmed by an AI-generated outcome and risks
undermining established frameworks for workplace justice. The spread
of AI also risks undermining the right to collective action and
collective bargaining. When AI enables individualised management
of workers and micro-targeted remuneration or performance regimes,
it may erode the solidarity on which worker organisation depends.
Ensuring social dialogue and meaningful worker representation as
AI is integrated remains an ongoing challenge.
6. Legal and ethical ambiguities around AI abound. The rapid
deployment of AI technologies often outpaces the ability of regulatory
frameworks to catch up, leaving significant grey areas regarding
the acceptable uses of AI in the workplace, the scope of worker
consent and the right to contest automated decisions. This regulatory
lag exposes workers to shifting standards and inconsistent protections.
Moreover, as workplaces become more reliant on interconnected AI
systems, they are exposed to greater threats from cyberattacks and
data breaches. For smaller enterprises and creative industries in
particular, as well as public services, the capacity to defend against
such threats may be inadequate, leaving intellectual property and personal
non-public data vulnerable to automated data scraping for unauthorised
uses.
7. AI systems in the workplace can have both positive and negative
effects on employees’ mental health. By automating repetitive tasks,
supporting decision-making and reducing workloads, AI can help alleviate stress,
improve efficiency and give employees more time for meaningful or
creative work. However, excessive monitoring, algorithmic management,
concerns about job security, reduced autonomy and the pressure to adapt
quickly to new technologies can contribute to anxiety, burnout and
feelings of inadequacy or isolation. The impact largely depends
on how AI is introduced and managed; organisations should therefore
involve employees in implementation decisions, provide adequate
training and support, maintain transparency about how AI is used
and ensure that technology complements rather than replaces human
judgment and workplace relationships.
8. Finally, there is the overarching risk of technological dependence
and “lock-in” within the public sector and at the societal level.
If AI systems, particularly those developed and controlled externally,
become embedded in core public or workplace functions, societies
may find themselves dependent on proprietary technologies, with
limited sovereignty and reduced ability to steer development or
ensure compliance with values such as transparency and equity.
9. The question thus is, are humans in control of AI tools in
the world of work and beyond? I should recall that the Parliamentary
Assembly’s Resolution 2345 (2020) “Artificial intelligence and labour
markets: friend or foe?” asked member States to ensure “sovereign
participation in and control of algorithmic developments, guaranteeing
the full respect for existing legal norms and standards by AI developers
and users in the context of employment, and avoiding regulatory
capture by influential AI businesses”. The same resolution asked member
States to design “a regulatory framework that promotes complementarity
between AI applications and human work, and ensures proper human
oversight in decision making”. (see proposed amendment D)
10. While the Council of Europe Framework Convention on Artificial
Intelligence adopted in 2024 requires signatory governments to adopt
national laws and regulations on AI lifecycles to ensure that AI
systems respect human rights, democracy and the rule of law, it
gives countries much latitude to apply these protections to private
sector businesses – or not. Yet private companies overwhelmingly
dominate the development of cutting-edge AI applications and their
use compared to public entities;
Note they should be held accountable
for the proper design and use of such applications at workplaces,
including the recruitment processes.
11. The AI genie is out of the bottle, and we have to live with
this reality. Policy makers must now heed the warnings issued by
major AI researchers and engineers shortly after disturbing revelations
about AI applications under development turning wild and uncontrollable,
beyond human comprehension, “escaping their test environments and
hacking into other systems completely autonomously” with a huge
destructive potential for critical infrastructure and humanity.
Note AI systems should work for
people, not vice versus. (see proposed amendment B)
12. In the circumstances, as rapporteur for opinion, I believe
that Council of Europe and OECD member States should step up their
regulatory action to strengthen the liability of all entities developing
and using AI modules and applications, notably in the employment
sphere, so as to protect human dignity, safety and occupational
health, and to ensure the responsible design and use of AI systems
in line with the requirements of the European Social Charter, the
European Union AI Act,
Note the
OECD Principles for responsible stewardship of trustworthy AI and
the related Due Diligence Guidance for Responsible AI, relevant
ILO conventions and national legislation. (See proposed Amendment
A)
