Case study / University of Zagreb / S02
Inside one finding.
The source article, the provisional result, and the reasoning that connects them.
Records as at . Assessed against StudentSafe Standard v0.1, a draft for consultation — the current rubric is v0.1.1, which changed only a publication rule. None of the 24 published findings has been independently verified.
One finding, opened up / S02
From an article to an assessment.
This is the University of Zagreb’s university-wide protection for a student who reports, as far as we could find it in its published rules. The assessment is provisional.
We askedAre students who report concerns, complain or act as witnesses expressly protected against retaliation, including academic retaliation?
Na osobu koja je odbila uznemirivanje ili prijavila takvu vrstu ponašanja ne smije se vršiti nikakav pritisak. TranslationNo pressure may be exerted on a person who rejected harassment or reported such behaviour.
- Safeguard
- S02 Explicit anti-retaliation protection
- Record
- University of Zagreb
- Article relied on
- Code of Ethics
Art. 12(5) - Document date
- Still in force?
- Not confirmed — 2023 Statute Art. 129(1)
- Language read
- Croatian
Why this is a 1 and not a 2.
The Standard asks the same question of every rule and grades the answer against written criteria. These are the two levels that were in play, in the Standard’s own words.
Level 1 — assigned
General whistleblower or anti-victimisation protection exists, but its application to students raising concerns about academic staff is unclear or indirect.
Level 2 — not reached
Retaliation against student complainants or witnesses is expressly prohibited, but academic forms of retaliation (grades, supervision, references) are not addressed, or no mechanism for reporting retaliation is specified.
Our reasoning
The protection is express and applies to students, but only to those who reject or report harassment; it does not reach complainants about other misuse of authority or witnesses, so the core of level 2 (protection of student complainants and witnesses) is only partly met. It does not address academic retaliation (grades, supervision, references), treat retaliation as a separate offence, or give a route to report retaliation. Currency caveat: this finding relies on the 2007 Code of Ethics, whose continued legal force after 1 October 2024 could not be confirmed (2023 Statute Art. 129(1)).
Where two levels are plausible, assign the lower level and explain the uncertainty in the rationale.
What this finding does not coverNo university-wide anti-retaliation rule was identified for student complainants or witnesses in general, and none that addresses academic forms of retaliation. The Gender Equality Plan's reference to this rule (section 2.5) is framed around employees.
How far this finding reaches We read the University’s university-wide rules. Faculty rules were not opened, and faculty-level harassment rules do appear to exist at some faculties, so this finding describes the university-wide layer only. National law also sits above it: Croatia’s Anti-Discrimination Act protects anyone who reports discrimination in good faith, which is assessed separately in the Croatia record.
The full criteria for S02, levels 0 to 3 This record, and the other eleven safeguards at this university
Where the sentence comes from.
Every finding names one document and the article inside it. This is the whole record behind the sentence above.
- Document
- Etički kodeks Sveučilišta u ZagrebuCode of Ethics of the University of Zagreb
- Issued by
- University of Zagreb (Senate)
- Document type
- Code of ethics
- Official reference
- KLASA: 012-03/07-03/2, URBROJ: 380-07/1-09-5
- Published
- Article relied on
- Code of Ethics Art. 12(5)
- Text status
- Official text, published by the University; the file is the 2009 print edition, and the 2018 amendments are published separately. Of the 17 sources in our register, 3 are unofficial consolidations rather than official gazette texts. This is not one of them.
- Accessed
- — open the PDF on unizg.hr
- Archived copy
- None — we have not archived a single one of the 17 sources yet, so if a document moves or changes, the version we read is gone.
- Register entry
- [hr-unizg-ethics-code-2007]
We cannot confirm this Code is still in force. It was adopted in 2007 under the University’s 2005 Statute, and Article 129(1) of the 2023 Statute required acts adopted under the previous Statute to be aligned by 1 October 2024. We found no re-adopted or consolidated Code, and the 2007 text is still published on the University’s regulations page. 10 of our 12 findings at this university cite it. Our own cross-check recommended writing to the University to ask whether it remains in force. That is still open.
Someone re-read every word. That is still not verification.
Researched with AI assistance from official sources, with an independent automated cross-check. All findings await verification by a human second reviewer.
On an independent automated cross-check re-downloaded every source cited in both Croatian dossiers, converted them to text and read every cited article verbatim, without using the researchers’ summaries. Some scanned documents were read as images, and one was only read in part. It agreed with 22 findings and asked for two to change.
- Confirmed
- 22 findings
- Adjusted
- 2 findings — Croatia S04 and S06, from “Unable to assess” to “Limited”. Both changes were made before publication.
- On this specimen
- “Code Art. 12(5) was verified verbatim, and the quotation is exact.” It also asked us to make the reason for the lower level explicit; the first sentence of the reasoning above is the wording it proposed.
An automated cross-check is not verification. Our own rubric reserves that word for something that has not happened: “A second reviewer independently checked the source, locator, interpretation and score against this rubric.” Nobody has done that for any of the 24 findings, and a reviewer may not verify their own work.
The full cross-check, in the repository How verification is supposed to work
What a result can mean.
The sentence above came out as “Limited”. A safeguard can come out six ways, and they are not six points on one scale — three of them are statements about our research rather than about an institution’s rules.
- Not yet researched
312safeguards
No research has been carried out for this safeguard in this jurisdiction or institution. Says nothing about whether a safeguard exists.
Nobody has looked. 26 Member States sit here, 12 safeguards each.
Research status by Member State - Unable to assess
1safeguard
Research was carried out, but the available sources were insufficient, inaccessible, contradictory or ambiguous, so no score can responsibly be assigned.
Croatia, written and reasoned decisions. The texts did not settle whether the national administrative-procedure rules reach complaints about staff conduct, and the plausible answers ran from 0 to 3.
Read that finding - Score 0: Not identified
2safeguards
Research was carried out against the listed sources and no qualifying provision was identified. This is not proof that no safeguard exists. A 0 is never final until a second reviewer has verified it. Until then it is published only as provisional, carries the complete list of sources reviewed, and must not be presented as a finding that the safeguard is absent.
Croatia S11, and the specimen university’s S05. Each one publishes every document that was read and found not to contain a qualifying provision — eight of them for S05.
Read the University of Zagreb one - Score 1: Limited
16safeguards
A relevant provision exists, but it is general, partial, discretionary, or lacks the operational elements needed to make it usable.
The commonest answer in this evidence base, and what the sentence above was graded.
- Score 2: Substantial
5safeguards
An explicit, operational provision exists and covers the core of the safeguard, but one or more elements required for "Strong" are missing.
The highest answer anywhere in this evidence base.
- Score 3: Strong
0safeguards
An explicit, operational, binding provision exists that meets every element listed for "Strong" for this safeguard.
No finding here reaches it. That is what the evidence base looks like today, not a conclusion about European higher education.
No finding is high confidence. 18 are medium, 6 are low.
Never average these, never add them up, and never turn a “not researched” into a zero. Not looking, looking and finding nothing, and establishing what a rule says are three different statements.
What a finding is not.
Not a place to report. We cannot take a complaint, act on one, or advise on a case, and we have no private channel. If something is happening to you now, start here.
We read documents. We do not observe practice. A rule can be well written and ignored, or thin on paper and handled well in person. A finding says what an institution has committed to in writing — nothing about how any real case was handled, and nothing about any individual.
Academic Safeguard Europe is not a court, inspectorate, regulator or disciplinary body. We do not decide whether anything happened, we do not assess individuals, and we do not rank universities or countries. There is no overall score, and there will not be one.
This is a research programme of StudentSafe Europe, a non-profit association based in Zagreb, Croatia. No funder, university, government or party decides a conclusion, changes an assessment, or sees a finding before it is published.
Tell us what we got wrong.
A document we missed, a newer version, a misread article, a broken link. Requests go through a public form on GitHub, and a pseudonymous account works. There is no private route yet. That is a gap, not a policy.
A request that raises a substantive question marks the finding disputed in public while someone other than its author reviews it, and nothing is taken down while that happens. A disagreement with the rubric itself is recorded and answered in the changelog rather than treated as an error.
Right of reply. Any institution with a profile, and any national authority whose framework we have assessed, may publish a statement of up to 300 words beside its findings, whatever those findings say. We publish it dated and unedited, removing only personal data or unlawful content. It does not give an institution editorial control, approval before publication, or the removal of a sourced finding. Before a new institutional profile goes up, we aim to write to the institution first.
No correction request has been received. The log is empty.
24 findings. No second reader.
Nobody has verified a single finding yet, and that is the number we would most like to be wrong about. Verifying one means opening the document we cited, finding the article, and saying whether the level is right. Native speakers of the language a document is written in are especially welcome.
Verify a finding