Codename EA20: When the Swedish Migration Agency's Language Analysis Becomes Hidden Evidence

SVT's investigation Codename EA20 revealed how the Swedish Migration Agency purchased language and knowledge analyses from Sprakab—while anonymous analysts, whose qualifications could not be verified, held decisive weight in asylum decisions.

Summary: When an asylum seeker lacks documents that make identity or residence likely, the Swedish Migration Agency may order a language and knowledge analysis. For a long time, this service was primarily purchased from the private company Sprakab. Analysts worked under code names – in several cases under the designation EA20 – and neither the applicant, the representative, nor practically even the authority could openly review who was making the statements, what education the person had, or how the audio file was actually assessed.

SVT's Uppdrag granskning revealed in 2014–2015 how this chain could become legally insecure: false credentials of a secret analyst, factual errors in knowledge questions, conclusions about "manipulation" that independent linguists dismissed, and decisions where the language analysis practically indicated which country should be specified in the deportation – despite the authority simultaneously claiming that the analysis never determines the case.

What a Language and Knowledge Analysis Really Is

The method is used when other investigative avenues are insufficient. The applicant speaks by phone with an analyst who asks questions about the hometown, institutions, geography, and everyday knowledge, while the language and dialect are assessed. The result is summarized in a report that the Swedish Migration Agency adds to the file.

On paper, it is a complement. In practice, the report can become the first – and sometimes only – clear indicator towards a country or region that the applicant has not themselves stated. When documents are missing, when the story is difficult to verify, and when the authority needs to specify a country in the decision, the external analysis fills a vacuum.

It is precisely this vacuum that makes anonymity dangerous. A regular expert in court can be heard, challenged, and tested. A code name analyst in a phone interview cannot usually be cross-examined in the same way. What remains is the company's summary – and the authority's interpretation of it.

Code Name EA20: The Secret Analyst

EA20 was highlighted by Sprakab as one of the foremost analysts in Somali. According to Uppdrag granskning, the person had been involved in nearly 5,000 language and knowledge analyses over approximately eight years. This is not a marginal sideline in a few cases; it is industrial volume in life-determining decisions.

With the help of British court documents, SVT was able to identify the man behind the code name and compare the resume with university records. Several courses and educational claims did not match. Independent Somali linguistics – including Professor Abdalla Omar Mansur at the University of Rome – also questioned the broad language skills that were highlighted. Additionally, information emerged about criminal convictions and debts with the Enforcement Authority, factors that according to the reporting were considered when the man himself applied for, but was denied, Swedish citizenship.

The then-legal director of the Swedish Migration Agency, Fredrik Beijer, told SVT that knowingly false information about competence is a threat to legal certainty and that the matter must be further investigated. Sprakab's CEO Pia Luu declined to answer questions about EA20 and the incorrect analyses in the program.

When the Report Omits What Speaks for the Applicant

UG compared audio files and complete question-and-answer sequences with Sprakab's concise reports. In several cases, parts that spoke in favor of the applicant were missing, while the analyst was actually wrong in knowledge questions.

In Mohamed's case – a man who claimed to have fled Al-Shabaab from Afgoye in southern Somalia – EA20's analysis pointed towards Somaliland. The conversation lasted just over 18 minutes. Mohamed correctly answered that the area has seven districts and listed them. In Sprakab's analysis, it stated that there should be ten districts and that Mohamed had failed. According to UG, the Swedish Migration Agency could have verified the information in its own knowledge bank.

For Jamac, who claimed southern Somalia, the analysis instead pointed towards Kenya or Tanzania – countries he claimed he had never been to. The pattern is the same: a short phone interview, an anonymous assessor, a summary that becomes significant in the decision, and an applicant who afterwards faces a conclusion that is difficult to challenge without access to the analyst's identity and competence.

The Bakr Family: Syria in the Story, Armenia in the Decision

A Yazidi case made the chain's effect particularly visible. The Bakr family claimed they came from northern Syria. The language analysis instead pointed out Armenia with high certainty. According to UG, Armenia was not in the investigation until the analysis established linguistic background there – and thereafter the Swedish Migration Agency decided on deportation there.

A university lecturer in Kurmanji at Uppsala University, who listened to the recording and met the family, questioned whether it is even methodologically defensible to draw such conclusions from an approximately 20-minute phone call. Armenia later informed that the family was not citizens of the country.

Press officer Fredrik Bengtsson simultaneously told SVT that a language analysis never has a decisive role and that it is only part of the total evidence. But in decisions UG reviewed, the formulation appeared that the language analysis is considered to have "a significant evidential value." When the authority's public message and decision texts diverge, it becomes difficult for the applicant – and for the public – to understand what role the secret analysis actually plays.

Sumayo, 13 Years Old: "Obvious Manipulation"

In the follow-up The Lying Child (2015), UG demonstrated that the same analyst EA20 was also used in cases involving unaccompanied minors. Thirteen-year-old Sumayo claimed she was from Buulaburde in southern Somalia. After the language test, Sprakab concluded that she attempted to imitate a southern Somali dialect she "did not master." A quick correction of the word for mosque was described as "deliberate manipulation"—a phrase later used in the Swedish Migration Agency's investigation.

Professor Mansur, who listened to the recording, instead assessed that Sumayo spoke a southern Somali dialect well and that the claim of imitation was unreasonable for a child of that age. The Children's Ombudsman criticized the occurrence of similar setups. The case highlights a particular rule-of-law risk: when an anonymous adult analyst attributes strategic linguistic manipulation to a child, and when the applicant cannot challenge the analyst's credentials, responding becomes nearly impossible.

The rule-of-law problem in four points

  1. Outsourced Evidence Evaluation. The state retains the decision but purchases the crucial origin assessment from a private company whose internal analysts are concealed.
  2. Asymmetric Party Position. The applicant cannot practically hear, cross-examine, or verify the person whose opinion is used against them.
  3. Discrepancy Between Policy and Practice. The authority states that the analysis is supplementary; decision texts and outcomes show that it can become the determining factor for identifying the country.
  4. Quality Control After the Scandal, Not Before. False credentials and factual errors were discovered through journalistic scrutiny and foreign court documents—not through visible, independent control that the applicant could rely on in real-time.

What Happened After the Revelation?

Following UG's reporting, several analyses were cross-examined, including by Dutch authorities; several were questioned. The Swedish Migration Agency had former Justice and Minister Sten Heckscher review the use of language analyses. He did not want to call the method legally insecure per se but criticized the quality and the excessive weight given to the analyses. Radio Sweden later reported that Heckscher, among other things, proposed clearer routines, coordination, and that analyses should not be ordered too early in the process.

In the parliamentary debate following the scandal, the government emphasized that procurement requirements, follow-up, and quality audits existed, and that both Sprakab and Verified AB were included in Heckscher's assignment. This is relevant as background—but it does not solve the fundamental question: how can an asylum seeker respond to an opinion when the expert's identity and credentials are secret, and when the authority itself sometimes lacks full insight into who actually made the assessment?

The Swedish Migration Agency's ethical council later, in a statement on automated language analyses (2020), warned that external language evidence is easily perceived as stronger than it is, and emphasized that language analyses should not have decisive significance to the applicant's disadvantage. The warning is essentially the same as UG's case illustrated with human, anonymous analysts: the technology or provider must not become a false answer key.

Why the Case is Still Relevant in 2026

Codename EA20 is a historical review, but the chain it exposed is not a museum. As long as asylum assessment relies on establishing origin without full documentation, the state will seek external methods: linguistic, technical, or commercial. Legal certainty is then determined by whether the applicant can understand, respond to, and have the evidence tested—not by whether the provider has changed name, method, or degree of automation.

For those involved in a migration case today, the lesson is concrete:

For practical work with timelines before oral investigation and correction afterward, also see the guides on preparation before investigation interviews and correction after oral investigation.

Conclusion

The EA20 scandal was not just about an individual analyst with false credentials. It was about a model where the Swedish Migration Agency outsourced a central part of the origin assessment to a private company, whose anonymous linguists could practically determine which country appeared likely—while the authority publicly downplayed the significance of the analyses.

A legally certain asylum procedure requires that evidence can be tested. When expert opinions are secret in their source, factually incorrect, and heavily weighted in decisions, legal certainty becomes a slogan rather than a method. This is why Codename EA20 remains a relevant warning system: not against language analysis as an idea, but against hidden, outsourced evidence without real party access to the evidence.

FAQ

What was Sprakab?
A private company that provided language and knowledge analyses to the Swedish Migration Agency. The analysts often worked under codenames and interacted with the applicant by phone.

What does the codename EA20 mean?
An internal designation for an analyst that Sprakab highlighted as particularly competent in Somali. SVT was able to show that the credentials were incorrect and that several analyses contained serious deficiencies.

Does the language analysis decide the asylum case?
The Swedish Migration Agency has stated that it never has a decisive role. In decisions and cases reviewed by UG, the analysis could still have "significant evidential value" and in practice indicate the country in the deportation decision.

Can the applicant find out who the analyst is?
In the model reviewed by UG, the identity was kept secret. This complicates the challenge of competence, bias, and methodological errors.

Is this legal advice in an ongoing case?
No. The text is a review of public sources and principle issues. Individual cases require their own assessment based on the file and applicable law.

Sources