Ref. Ares(2022)8519438 - 08/12/2022
Summary of responses to the questionnaire on prioritisation of AI
use-cases
Total number of responses received: 12
This document provides a summary of responses to the questionnaire and indicated the use-cases
prioritised based on these responses. Number of points to calculate priority scores was calculated on
the basis of the total number of use cases and priority order in which those were indicated. For
example: If there are a total of three use cases, then the highest priority use-case received 3 points,
second – 2 points, last – 1 point. If there are six use-cases, then the highest priority received 6
points, second-highest – 5 points, last – 1 point. These points were then summed and the use-case
receiving the highest number of points was prioritised.
Short stay visa issuance
One MS indicated that ‘Visa application triaging has been implemented’.
Other respondents indicated that none of the use-cases have been implemented at MS level.
Use case
Priority scores
Priority #
Application triaging
25
2
Identification of irregular travelling patterns
28
1
Tailored application form
19
3
Issuance of long-stay visas or residence permits in the Schengen area
Four respondents indicated that one of the use cases was implemented at MS level:
-
Virtual assistant supporting long-term stay permit/migration application process (Two MS)
-
Automatic triaging of applications to speed up risk assessments (Two MS)
Use case
Priority scores
Priority #
Application chatbot
20
2
Application triaging
34
1
Moving within the Schengen area
18
3
Asylum
None of the use-cases have been reported as implemented at MS level.
Use case
Priority scores
Priority #
Vulnerability assessment
43
1
Registration chatbot
34
4
Abscondment risk assessment
40
2
Refugee allocation
23
5
Intelligent search engine
40
2
SIS/SIRENE
Five MS reported national implementation of two cases:
-
Alert detection
-
Automatic form completion (2 MS)
-
AI tool to aid in the knowledge management of SIS (2MS)
Use case
Priority scores
Priority #
Alert detection
19
3
Knowledge management/search tools
23
2
Automatic form completion
30
1
Transversal processes
Several respondents indicated certain use-cases implemented:
-
Detection of forged travel documents (Six MS);
-
Identification of fraudulent supporting documents (Three MS);
-
Improved biometric matching (facial recognition) (Four MS);
-
Consistent decision-making (historical case reasoning engine);
Use case
Priority scores
Priority #
Multi-lingual translation
56
3
Identification of fraudulent supporting documents
62
1
Consistent decision making (historical case reasoning 54
5
engine)
AI to monitor the ethicality of other AI systems
21
7
Detection of forged travel documents
61
2
Post-application monitoring of TCNs
27
6
Improved biometric matching (facial recognition)
55
4
Other use-cases suggested by respondents
-
Transcription (speech to text) - this is in implementation in some systems
-
Categorization of criminal data for use of analysing.
-
for visa applicants, there are a lot of requirements for supporting documents, appendixes for
visa application form. It would be good to have a database where all the supporting
documents are scanned and anonymized. And AI would notice any pattern common to
fraudulent purpose of visa applicant. The data pool for supporting documents is like big data
where all kind of research and machine learning can be possible depending on need and
possible scenarios.
-
Other use case is for issuing authority information in travel documents. For instance it is
useful for decision makers and authorities that certain issuing authority with certain
passport number may need additional investigation in order to find out the real purpose of
visa applicant. Thus, the information about issuing authority is important (but unfortunately
not in use any more)
-
Additionally use case may be, that if somewhere someone has stolen a bunch of passport or
empty visa stickers, the numbers or series of those passport numbers can be checked
against database when dealing with visa applications. AI could help to resolve whether the
visa applicants passport belongs to the certain number of passports.
-
Implementation of AI assisted filling of SIRENE forms is projected in the near future
Implementation of use cases
1. Seven out of 12 respondents indicated that they would be interested in collaborating on
the prioritised use-cases.
2. The following use-cases were indicated by the MS interested in collaboration
a. One of the challenges in LEA is analysis of child sexual abuse material. Investigators have
to find relations in large amounts of seized images, videos and sounds. Without AI
techniques this is not possible. These AI techniques can detect object in these images
and videos, recognize similar faces, classify similar material in the same group etc and
can be provided by eu-LISA as well as repository for large amount of these material.
Without this centralized solution each member state has to find own AI components and
set own data repository for this material.
b. Visa related project with MS with similar privacy regulation.
c. Identification of fraudulent visa applications based on the analysis of supporting
documents.
d. Identification of forged passports/visas based on passport/visa number (stolen blank
passport books/Visa stickers)
e. Implementation of AI assisted filling of SIRENE forms
3. 11 out of 12 respondents indicated that they’d need additional funding for the
implementation of specific use-cases
4. Nine out of 12 respondents would see a role for eu-LISA in the implementation of
prioritised use-cases. The roles indicated are:
a. Provide centralised infrastructure (e.g., compute);
b. Facilitate collaboration with authorities from other Member States
c. Provide data sets for model training/testing;
d. Help to create a regulatory environment to secure the legal compliance of the pilot.
e. Provide expertise in AI, project lead, coordinator
