Observations on the Draft Law “Implementation of the Pact on Migration and Asylum” regarding the use of Artificial Intelligence in asylum procedures

On 11 May 2026, the Ministry of Migration and Asylum launched an online public consultation on the draft law entitled “Implementation of the Pact on Migration and Asylum and other provisions of the Ministry of Migration and Asylum.” 

The draft law introduces a new framework for screening procedures, international protection processes, the operation of Eurodac, and the management of biometric and personal data of asylum seekers and migrants. It explicitly provides for the possibility of using artificial intelligence systems in the provision of information, interpretation, translation, and the recording of personal interviews of applicants for international protection. 

These provisions raise serious concerns regarding data protection, transparency, non-discrimination, and effective judicial protection in light of the General Data Protection Regulation (GDPR), the Charter of Fundamental Rights of the European Union, and the Artificial Intelligence Act (AI Act).

The draft law establishes an extensive framework for the digital management of asylum and migration procedures, including:

• Collection and matching of biometric data
• Operation of interoperable information systems
• Use of artificial intelligence systems in informing asylum seekers, as well as in translation and interpretation during asylum procedures
• Use of artificial intelligence systems for audio recording and AI-generated asylum interview transcripts

Particularly critical provisions include:

• Article 8 (“Registration of personal data and collection of biometric data”)
• Articles 243–257 concerning the Eurodac system and biometric data matching
• Article 96(3) and (5) on the use of AI-based information systems for information provision, automated translation, and subtitling
• Article 101(1) on the use of AI systems for automatic transcription of personal interviews

Although the draft law does not explicitly provide for “profiling” or “risk assessment,” the systematic collection and interconnection of biometric and personal data, combined with automated matching mechanisms and the use of AI in information provision, translation, subtitling, and interview recording, effectively creates the technical and institutional infrastructure for automated categorisation and assessment of applicants for international protection.

The use of automated translation software and AI-assisted interpretation in asylum procedures raises serious concerns regarding reliability and the protection of fundamental rights. Asylum procedures rely on precise linguistic representation; applicants often belong to vulnerable groups, particularly unaccompanied minors who are entitled to additional procedural safeguards; and credibility assessments depend on subtle linguistic, cultural, and contextual nuances. 

AI-based interpretation removes the human element, is incompatible with practices that take trauma into account, and cannot ensure a safe and supportive environment for individuals who have experienced violence, persecution, or other traumatic events.

Similarly, Article 101 provides that when AI systems are used to generate asylum interview transcripts, no other transcript is required. This undermines essential procedural safeguards, as it does not ensure human oversight of the asylum interview transcript, does not guarantee a right to correction in case of errors resulting from AI use, and does not provide safeguards against errors in speech-to-text systems.

Furthermore, the draft law does not include specific safeguards regarding the transparency of algorithmic systems, error rates, independent certification, accuracy audits, the conduct of fundamental rights impact assessments for the use of AI systems, or the right to object or opt for a fully human procedure.

These provisions must be assessed in light of Article 22 of the GDPR, Articles 7, 8, 21, and 47 of the EU Charter of Fundamental Rights, as well as the AI Act, which classifies AI systems in the field of migration and asylum as high-risk.

For these reasons, the following are considered necessary:

• Explicit prohibition of the use of AI systems for credibility assessment or biometric categorisation
• Prohibition of “emotion recognition systems”
• Mandatory human oversight and validation of all asylum interview transcripts produced or assisted by AI systems, as well as AI-generated translations
• Ensuring the right of access and correction
• Establishing strict safeguards of necessity, proportionality, and transparency for any processing of biometric data in the context of asylum and migration procedures

In conclusion, the introduction of AI systems in interview processes, translation, and transcription, combined with extensive biometric data processing, creates the technical possibility for developing practices of automated credibility assessment, biometric categorisation, or emotion analysis without explicit legal prohibitions or adequate safeguards.

Signatory organisations:

• I Have Rights
• Network for Children’s Rights
• WHEN Equity Empowerment Change
• Homo Digitalis
• Open Lab Athens
• Greek Forum of Migrants
• HIAS Greece