Artificial intelligence and migration
- Author(s):
- Parliamentary Assembly
- Origin
- Assembly debate on 3 October 2025
(36th sitting) (see Doc. 16240,
report of the Committee on Migration, Refugees and Displaced Persons,
rapporteur: Mr Petri Honkonen). Text
adopted by the Assembly on 3 October 2025 (36th sitting). See
also Recommendation 2300
(2025).
1. The Parliamentary Assembly acknowledges
the transformative potential of artificial intelligence (AI) across
a range of sectors, including migration management. AI systems –
capable of autonomous decision making and complex data analysis
– are increasingly used in border surveillance, visa processing,
biometric identification, natural language processing and integration
support. These applications promise enhanced efficiency and service
accessibility for migrants, refugees and asylum seekers.
2. The Assembly also underlines the opportunities created by
the responsible and ethical use of AI. AI can contribute to saving
lives through enhanced search-and-rescue operations, provide migrants
and refugees with multilingual and accessible information and support
their integration by facilitating language learning, education and
access to labour markets. When designed with transparency, human
oversight and strong human rights safeguards, AI can foster innovation,
reduce administrative burdens and strengthen trust in institutions.
In this way, AI can make migration management not only more efficient
but also more humane and inclusive.
3. The Assembly emphasises, however, that technological innovation
must not come at the expense of fundamental rights. If used improperly,
AI can reinforce structural inequalities, infringe on privacy and undermine
asylum protection. The Assembly therefore reiterates its call for
all Council of Europe member States to sign and ratify the Council
of Europe Framework Convention on Artificial Intelligence and Human Rights,
Democracy and the Rule of Law (CETS No. 225, “Framework Convention
on AI”), aiming to ensure the development and deployment of AI in
line with human rights standards and explicitly prohibiting AI applications that
violate the right to seek asylum.
4. AI-driven modernisation must be carried out in a way that
minimises dangerous effects and risks for migrants, refugees and
asylum seekers, such as discrimination and bias, and that does not
unintentionally reinforce existing stereotypes or prejudice. States
should instead harness the potential of AI to foster a more inclusive,
secure and humane migration system.
5. Recognising the profound impact that AI can have on individual
rights and liberties, the Assembly stresses that AI should support
– not replace – human decision making in migration and asylum processes, even
if in some cases AI may offer greater security and effectiveness
than human decision making alone, by reducing the scope for human
error. All AI tools must be transparent, reliable and subject to
oversight. They must also be deployed in alignment with key international
instruments, including the European Convention on Human Rights (ETS
No. 5, “Convention”), the 1951 United Nations Convention Relating
to the Status of Refugees as amended by its 1967 Protocol (“Refugee
Convention”) and the Charter of Fundamental Rights of the European
Union.
6. Noting that the European Union Artificial Intelligence Act
classifies migration-related AI as high-risk technology, the Assembly
underscores the need for additional safeguards. The use of AI in
migration, asylum and border control management must not allow for
the bypassing of international obligations, in particular under the
Refugee Convention. Nor should it be used in any way to infringe
on the principle of non-refoulement,
or to deny safe and effective legal avenues into States’ territory,
including the right to international protection.
7. It is important therefore that AI systems in migration and
asylum procedures undergo human rights, democracy and rule of law
impact assessments before their deployment. The Assembly recommends
the use of the Council of Europe methodology for the risk and impact
assessment of artificial intelligence systems from the point of
view of human rights, democracy and the rule of law (HUDERIA) to
identify and mitigate risks, including algorithmic bias and privacy
violations. Oversight must be embedded throughout the life cycle
of AI systems with independent evaluations and mandatory human review.
8. The Assembly calls for a prohibition on the use of AI tools
such as automated credibility assessments, emotion recognition and
risk profiling based on nationality or ethnicity. These technologies
lack scientific validity and are incompatible with Articles 3 and
14 of the Convention.
9. Noting the critical importance of data protection, privacy
and security in the use of AI for asylum procedures, the Assembly
emphasises that sensitive data, including biometrics, interview
transcripts and country-of-origin information, must be end-to-end
encrypted and must not be shared with the country of origin in cases
of risk of persecution, in accordance with the data protection policy
of the Office of the United Nations High Commissioner for Refugees
(UNHCR).
10. AI applications in migration management must thus balance
efficiency gains with rigorous protection of human rights at all
stages of the migration journey. The specific issues related to
the use of AI in the field of migration must be addressed for each
distinct stage of the process, from pre-departure activities to
transit, arrival, stay, circulation, temporary or permanent return
and sustainable (re)integration. Priority should be given to safeguards
against bias and human oversight, and to ensuring the dignity of
migrants, refugees and asylum seekers, as follows.
11. At the pre-departure stage, AI tools used in visa and asylum
screening should undergo human rights impact assessments before
deployment. Systems like the European Travel Information and Authorisation System
(ETIAS) should provide for manual review in cases flagged as high
risk. All screening tools must be transparent and free from discriminatory
outcomes. AI-powered chatbots can provide real-time multilingual information
to migrants, refugees and asylum seekers, provided they disclose
their automated nature and comply with ethical standards to avoid
spreading misinformation.
12. As regards labour migration via the appropriate channels,
AI tools may be used to streamline, in a fair and accessible manner,
the application process for employment and residence permits. AI
may further support the detection of exploitative recruitment practices
by identifying predatory employers and alerting the relevant authorities.
AI tools, such as visa checkers, should be audited for bias and
provide transparent outcomes that can be contested.
13. The Assembly encourages the ethical use of AI in climate migration
forecasting through the analysis of environmental and socio-economic
data to improve humanitarian planning in accordance with its
Resolution 2401 (2021) “Climate
and migration”. AI tools, such as those developed by the International Organization
for Migration (IOM), can provide policy makers with a clearer predictive
picture of climate change-driven displacement patterns. Such predictive
models can inform proactive support measures.
14. AI may also be used to identify and disrupt human smuggling
networks through responsible data analysis.
15. During the transit stage, the Assembly stresses the need for
strict regulation of AI surveillance technologies. Predictive policing
and biometric identification must be limited to strictly necessary
and proportionate use, and all biometric data must be encrypted
and restricted to authorised personnel. The Framework Convention
on AI limits biometric categorisation to strictly necessary and
proportionate cases. AI systems like the European Union’s European
Border Surveillance System (EUROSUR), which monitor border movements,
must comply with the IOM’s 2023 Data Protection Principles.
16. Surveillance technologies for migrants, refugees and asylum
seekers in transit must be regulated with a view to protecting their
rights, and AI-driven mass surveillance, such as drones and facial
recognition, should be prohibited. AI may be used positively to
support humanitarian corridors by using conflict mapping to identify safer
transit routes.
17. The Assembly recognises the essential role of the European
Union Agency for the Operational Management of Large-Scale IT Systems
in the Area of Freedom, Security and Justice (eu-LISA) and other European
Union bodies in managing migration-related databases such as the
Schengen Information System (SIS), the Visa Information System (VIS)
and Eurodac (the European Union’s centralised biometric database). Co-operation
between European Union and non-European Union Council of Europe
member States is important to ensure consistent data protection
standards and respect for humanitarian obligations.
18. The Assembly expresses once more its deep concern and sorrow
over the deaths at sea of migrants, refugees and asylum seekers.
It calls on member States to employ AI technologies to enhance search-and-rescue
capabilities and to uphold the dignity of deceased individuals,
in accordance with the principles laid down in its
Resolution 2569 (2024) “Missing
migrants, refugees and asylum seekers – A call to clarify their
fate” and its
Resolution 2612
(2025) “Saving the lives of migrants at sea and protecting
their human rights”.
19. At the arrival stage, AI can be used to support individualised,
fair and rights-compliant asylum procedures, while never replacing
the role of human caseworkers in interactions and decision making.
AI-generated documents should be accessible in applicants’ languages
and in plain language formats.
20. Biometric systems and language recognition tools should be
subject to routine bias audits to ensure equitable treatment, while
facial recognition tools, such as those used in smart border tunnels,
should undergo demographic testing to ensure compliance with non-discrimination
requirements and adhere to European Union-wide audit standards.
21. AI systems in asylum processing must be verified and corrected
for skewed datasets to avoid discriminatory proxies, and AI-generated
evidence must be verified by humans, with access to judicial review. The
use of tools such as emotion recognition or lie detectors must not
be included and predictive tools assessing the likelihood of absconding
must not be used to justify detention, especially of minors. Impact assessments
such as the HUDERIA methodology must be carried out prior to any
roll-out of new systems.
22. During the stay period, inclusive integration policies, in
accordance with
Resolution 2502
(2023) “Integration of migrants and refugees: benefits
for all parties involved”, can be supported by the ethical use of AI,
which can play a key role in accelerating the self-reliance of migrants,
refugees and asylum seekers, and in boosting the resilience of host
communities. Tools may be co-developed with refugee communities
and non-governmental organisations (NGOs) and made accessible offline
(telephone hotlines without internet access for example) and through
voice interfaces to bridge digital divides.
23. AI labour-matching tools should prioritise ethical criteria
such as family unity and cultural fit. Gender-sensitive design is
essential to avoid reinforcing labour market segregation that directs
women into low-wage sectors. Continuous feedback loops should help
address misplacements. Annual audits of AI systems used in welfare
and housing are essential to detect and correct bias. Predictive
analytics can support equitable urban planning, helping to prevent
segregation and foster innovation and safety in diverse communities.
24. The Assembly calls for the creation of independent oversight
bodies, such as the European Artificial Intelligence Board, which
should include civil society and legal and technological experts,
and should monitor the implementation of AI-powered systems in migration,
asylum and border control management based on the guidelines of
the UNHCR, the Framework Convention on AI and relevant European
Union regulations.
25. Robust redress and compensation mechanisms must also be available
to allow for the contesting of AI-generated evidence through expedited
legal channels. Legal aid should be expanded to cover algorithmic disputes.
26. For circular migration and return stages, AI chatbots used
in return support programmes should comply with fairness standards,
avoid nudging techniques and provide unbiased information. Migrants,
refugees and asylum seekers should retain control over their data,
with the ability to delete or transfer information upon exiting
programmes and a guarantee against the sharing of biometric information
with the country of origin. The sharing of biometric data with countries
of origin must be banned if there is any risk of persecution. AI-based assessments
of environmental conditions may assist in determining the safety
of return destinations.
27. Sustainable reintegration requires robust post-return monitoring.
States should implement community-informed impact assessments and
ethical AI tools to track outcomes related to employment, housing
and well-being. Offline accessible AI assistants must support migrants
to navigate reintegration services.
28. In line with its
Resolution 2343
(2020) “Preventing discrimination caused by the use of
artificial intelligence”, the Assembly underscores the importance
of specific action to prevent discrimination and disproportionate
negative impact on groups such as women, minorities and the most
vulnerable and marginalised individuals, including migrants, refugees
and asylum seekers.
29. Such potential discrimination in AI should be addressed at
the design phase, which can usefully benefit from the participation
of civil society organisations representing migrants, refugees and
asylum seekers to increase trust and reliability. Awareness raising
and training for asylum officers, persons employed by NGOs and AI
developers will strengthen ethical AI deployment.
30. Member States must safeguard against the misuse of AI for
disinformation, manipulation or cyberattacks that exploit migration
vulnerabilities. Geopolitical threats, including those linked to
the Russian Federation’s aggression against Ukraine, demand increased
vigilance and resilience in migration policy frameworks to avoid a
negative impact on the protection offered.
31. As regards the general climate of opinion around migration
issues, States should ensure that AI-powered chatbots are used ethically
to provide accessible, accurate and multilingual migration information,
and to ensure that these tools are not used to manipulate narratives
or asylum decisions. Here too, participatory design in chatbot development
is to be favoured.
32. To support the above measures, awareness raising and capacity
building among all public and private stakeholders – particularly
public authorities and officials, developers, small and medium-sized
enterprises and start-up AI enterprises – on the use of AI in migration
management are essential, based on the relevant regulatory frameworks
and practical implementation.
