Prison file after the amnesty: why it should be cancelled (and the RisCanvi risk)

By Mariam Bataller, lawyer and criminologist.

The Catalan independence (procés) amnesty has ordered the deletion of the criminal records of those convicted, but Organic Law 1/2024 says nothing about their prison file or about the personality protocols and risk assessments generated during their imprisonment. This article defends a specific thesis: most of those records should be cancelled, that is, removed from the operational circuit of the prison administration, and not merely blocked.

The decisive argument is not only about privacy: it is functional. The algorithmic risk-assessment systems already operating in Spanish prisons, with Catalonia’s RisCanvi as the central example, rely precisely on the kind of information a prison file contains. As long as that data remains available, a future entry into prison, even for facts unrelated to the procés, could reactivate the old profile and give the amnestied facts a practical capacity to influence decisions. The amnesty should have closed that door.

A half-way amnesty: the prison file was left out

In October 2026 the judicial cycle of the procés came to a close: the Constitutional Court upheld the amnesty for embezzlement, and the Criminal Chamber of the Supreme Court declared the criminal liability of the convicted leaders extinguished, lifted their disqualifications and sent the cancellation of their criminal records to the Ministry of Justice.

Organic Law 1/2024 was meticulous in ordering oblivion. Article 4.f provides that “criminal records arising from the conviction for the amnestied offence shall be deleted”; Article 5.2 orders the deletion of offences from the Central Register of Offences against Public Safety; Article 6.3, the unfavourable entries in public employees’ service records. The legislator went through, one by one, the registers capable of reproducing legal consequences of the amnestied act.

And yet it stopped short of the richest register of all. The law does not mention, either to order its deletion or to authorise its retention, the prison file: the set of documents reflecting an inmate’s life during imprisonment.

What a prison file contains

This is no minor omission. The file holds practically everything the prison administration knows about a person deprived of liberty: their procedural status, conduct, communications, disciplinary sanctions and, above all, the personality protocol. The General Prison Act requires one to be drawn up for every convicted person (art. 15.2), with its “scientific study of the personality”, the “determination of the criminological type” and the “diagnosis of criminal capacity and social adaptability” (art. 15). The Prison Regulations describe it as a “scientific study of the constitution, temperament, character, aptitudes and attitudes of the subject” recorded “in the inmate’s protocol” (art. 112), whose creation and custody they entrust to the Treatment Board (art. 273.n).

Closing the criminal-records register is not enough if the profile remains operational.

Keeping the file is not neutral: it is a mortgage on the future

Against cancellation it is often argued that the prison file is an administrative and historical document, protected by documentary heritage rules, and that keeping it harms no one as long as no one consults it. This reasoning rests on a questionable premise: that stored data is inert. In today’s Spanish administration, stored data is data waiting for a tool to claim it. And prison files contain precisely the kind of information those tools consume.

RisCanvi: the algorithm fed by the prison file

Since 2009, the Catalan prison system has classified its inmate population, including those on parole, using the RisCanvi risk-assessment protocol, a pioneering instrument in Spain that the Generalitat presents as support for professional judgement in “classification and the granting of leave” and in treatment planning. In other words, it influences decisions such as third grade, prison leave or parole.

The full scale weighs 43 risk factors; the available descriptions include educational level, history of violence, mental health, addictions and hostile attitude, and the initial screening scale includes variables such as “prison conduct problems”, “limited response to psychological or psychiatric treatment” or “hostile attitude or pro-criminal values”. The protocol’s official documentation recommends gathering information by cross-checking “interviews, files, coordination with other professionals”.

Two points of rigour. First: RisCanvi is formally a decision-support tool for professionals, not an autonomous machine; but its role is structural, because it frames the technical debate on classification, leave and intervention. Second: Instruction 2/2019 of the Secretariat for Criminal Measures sets up a follow-up procedure for released inmates assessed as being at high risk of violent reoffending for certain offences; its scope is limited, but it illustrates how far the system’s logic reaches: risk assessment does not end at the prison gate.

The risk: the old profile weighing again

If in the future one of the amnestied persons were to enter the prison system again for any reason, with no connection whatsoever to the procés, nothing today verifiably guarantees that the old file and assessments would not be retrieved for the new risk assessment. We argue that the legal order must prevent that reuse through express safeguards and, where no legitimate purpose for retention remains, through deletion. The old data would be read not historically but predictively: a conduct entry from 2019 could weigh again in a decision in 2030. That is exactly the risk that the deletion of records in Article 4.f seeks to avert, and that keeping the file leaves open.

A questioned algorithm: the RisCanvi audits

The urgency is not theoretical: RisCanvi’s reliability is publicly disputed. The audit commissioned by the Department of Justice from Dribia (January 2024) concluded that the algorithms show good overall performance and no serious discriminatory bias, but detected mismatches in subgroups and recommended publishing the code and documentation. A later adversarial audit by Eticas (2024) found “substantial deficiencies” in the system’s reliability and “opaque” calculations; El País summed it up in a headline as an algorithm that “works randomly”. Two audits, two methodologies, one thing in common: opacity and doubts about reliability.

RisCanvi is not the only tool. The University of Valencia has jointly studied RisCanvi, VioGén (gender-violence risk) and VeriPol (AI detection of false reports) from a fundamental-rights perspective. We do not claim that all of them consume prison files, but they share a logic: administrative decision-making assisted by data. Every record the administration keeps is a potential input to that logic. Deleting data is no longer just a matter of privacy: it is a matter of control over future decisions.

What European law says about retaining prison data

This thesis is not a mere assertion. It rests on three levels of support.

1. Judgment C-118/22 of the Court of Justice of the EU

The judgment of 30 January 2024, Case C-118/22 (ECLI:EU:C:2024:97), examined a national rule that retained the criminal and biometric data of convicted persons who were later rehabilitated until their death, with no possibility of erasure. The Court held it contrary to the principles of data minimisation and storage limitation in Directive (EU) 2016/680: a rule that does not require the controller to periodically review whether retention remains necessary, and does not recognise a right to erasure once data are no longer needed, is disproportionate. The ruling does not directly decide the fate of prison archives after an amnesty, but it offers a relevant interpretive criterion: retention after criminal liability has been extinguished requires specific and reviewable justification, not perpetual archiving.

2. Organic Law 7/2021 on data protection in criminal matters

Spanish law already incorporates that standard. LO 7/2021, which applies to the enforcement of criminal penalties and to inmates’ data (art. 15 bis of the General Prison Act), requires in Article 8 that data be kept “only for the time necessary” for their purpose, with review at least every three years, and in Article 23 erasure within a maximum of one month when processing is unlawful or there is a legal obligation to erase. The purpose of enforcing the amnestied sentence has ended and, with it, the basis for actively retaining the file as a management tool. Other purposes (evidentiary, welfare, archival) may remain for specific data, but they must be identified, delimited and time-limited, not used as an umbrella to keep everything.

3. The European Artificial Intelligence Act

Regulation (EU) 2024/1689 prohibits in Article 5(1)(d) the use of AI systems to assess or predict the risk of a person committing a criminal offence based solely on profiling or on assessing their personality traits and characteristics, except for systems that support human assessment based on objective, verifiable facts directly linked to a criminal activity. The provision does not prohibit keeping personality data, nor does it allow RisCanvi to be labelled a prohibited practice without further analysis. But it does codify a European distrust of predicting criminality from a psychological profile. Having historical information and being entitled to use it in a predictive assessment are different things, and the prison personality protocol is, by definition, an administrative psychological profile.

Amnesty and pardon: why they do not have the same effects

The procés leaders were pardoned in 2021 and amnestied in 2024-2026, and the two measures are not equivalent. A pardon forgives the execution of the sentence but does not erase the conviction: that is why its 2021 beneficiaries kept their records until the amnesty, and nothing follows from a pardon about deleting records or the prison file. An amnesty, by contrast, extinguishes criminal liability and, with LO 1/2024, entails the deletion of records (art. 4.f).

In both scenarios the need to keep profiles and risk assessments can be questioned, but not with the same legal automatism. The cancellation thesis rests on the amnesty and its systematic logic; for the pardon period (2021-2024), what applies is the necessity review that LO 7/2021 requires for any processing, not automatic deletion. The amnesty is the legal door through which the argument enters; the pardon explains why those files already contained years of follow-up data from semi-open regime.

The objections to cancellation and why they fall short

An honest lawyer must acknowledge the opposing views. There are three, and they deserve an answer.

“The file is a public document with historical value”

It is true that Law 16/1985 on Historical Heritage forbids destroying public documents while their evidentiary value remains (art. 55.2) and that deletion requires an authorised procedure; it is also true that the Higher Appraisal Commission for Administrative Documents has appraised state prison series (personal files, identification cards, admission books) for preservation in historical archives. Catalonia has an equivalent system of appraisal tables approved by the CNAATD, although we have not identified the specific table for files from Generalitat prisons; any conclusion on their archival fate must therefore remain conditional on that appraisal.

But the objection confuses two levels: the document and the active data. That a document has historical value does not require personal data to be kept in prison management circuits. Historical interest is satisfied by a restricted-access or anonymised archive copy, provided anonymisation does not strip the document of the value to be preserved. Algorithmic risk, however, is not mitigated by historical access restrictions, because the future user of the data would not be the historian but the prison system itself.

“The legal order already provides for separate retention”

Royal Decree 95/2009 keeps cancelled entries in a special section for judges (art. 19.3). This is the most serious counter-argument: it seems to show that the Spanish model of oblivion is always segregation, never destruction. But the analogy fails on the decisive point. The special section of the Criminal Records Register is reserved for judges and courts, for specific jurisdictional purposes; the active prison file is a tool of everyday administrative management and a potential input for algorithmic assessment. Transferring that model to the operational prison file would turn a jurisdictional exception into an administrative rule. And even if segregation is accepted, it would justify keeping basic identification, not the personality protocol or the risk assessments.

“The amnesty does not erase the historical fact of having been in prison”

Correct, but it does not settle the debate. No one proposes rewriting history or denying the imprisonment on paper: the aim is to prevent the documentary account of an amnestied act from remaining an input available for future administrative and algorithmic decisions. Pending evidentiary value (compensation, state liability, benefits) is protected by temporarily blocking the strictly necessary data, not by indefinite active retention.

Delete, block or keep: what should happen to the file

The debate is resolved by distinguishing the institutions properly:

• Blocking (art. 32 LOPDGDD, for processing subject to its regime): keeps data only to determine liabilities and destroys it once those lapse.
• Restriction of processing (LO 7/2021): restricts use in the cases set out in Article 23.
• Deletion: removes the data from processing when its conditions are met.
• Archival retention: separates the data from management and places it under the archives regime.

A legally correct blocking does not by itself authorise predictive reuse of the data. That is why the formal objection “it is already blocked” falls short: calling a file blocked is not enough if it remains available for building profiles, and when no purpose legitimises keeping the data and no evidentiary or archival obligation applies, indefinite blocking is not enough either: deletion is the proper course.

Our proposal: effective cancellation

We argue for the deletion from the operational circuit of most components of the amnestied persons’ files: first, the personality protocol, the risk assessments and the conduct entries; second, the monitoring and treatment records. What would remain, where appropriate, is basic identification (which the Generalitat’s own notice on prison activity treats as residual data after blocking) and restricted-access archive copies guaranteeing historical and evidentiary value, duly appraised.

Article 9.2 of the Prison Regulations, which allows data not to be cancelled for “reasons of public interest, security and protection of the rights of third parties” or for its “historical and statistical value”, requires a balancing of interests. And that balancing, carried out after an amnesty that has extinguished criminal liability, the enforcement purpose and the criminal records, can hardly conclude that the public interest lies in keeping intact the criminological profile of people whom the State has formally renounced treating as offenders. The advantage of deletion over blocking is not symbolic but structural: it is the only guarantee that does not depend on the future good behaviour of any computer system.

Conclusion

An amnesty is, in essence, an institutional declaration that certain acts will cease to produce effects. A legal order that deletes the criminal record but keeps the operational profile has not fully kept its promise. And in the age of algorithms, that profile does not sleep: it waits, with its 43 factors ready, for someone to read it again.

Frequently asked questions

Does the amnesty law delete the prison file?

Not expressly. LO 1/2024 orders the deletion of criminal records (art. 4.f), offences in the Central Register of Offences against Public Safety and unfavourable entries of public employees, but it does not mention the prison file or the personality protocol.

What is RisCanvi?

It is the risk-assessment protocol used by the Catalan prison system since 2009. It weighs 43 factors and supports decisions on classification, leave and treatment. Its information comes, among other sources, from interviews and files.

What is the difference between blocking and deleting data?

Blocking keeps the data under restricted access, only to determine liabilities, until they lapse. Deletion removes the data from processing. If there is no longer a legitimate purpose or obligation to keep it, deletion is the proper course.

Can the cancellation of prison data be requested?

LO 7/2021 recognises the right to erasure when processing is unlawful or there is a legal obligation to erase, and requires periodic review of whether retention remains necessary. Each case needs specific analysis. If you want to assess yours, you can contact our prison law firm.

---

Author’s note: this article is doctrinal and educational in intent, not legal advice. It distinguishes between what is verified (the texts of LO 1/2024, the General Prison Act, the Prison Regulations, LO 7/2021 and RD 95/2009, judgment C-118/22 of 30 January 2024, AI Act 2024/1689, the official RisCanvi documentation and its audits) and what is defensible interpretation: the extensive reading of art. 4.f of the amnesty law, the balancing under art. 9.2 of the Prison Regulations after the amnesty, and the application of the C-118/22 doctrine to prison files. Part of the material on RisCanvi comes from extracts of official documentation and press, cited in the text. The author has not found any administrative or judicial decision that has yet applied the amnesty to specific prison files, nor the specific Catalan appraisal table for those document series: both questions remain open and this article aims to contribute to the debate.

Are you affected by a risk assessment, a classification or a prison file? Our team of prison lawyers in Barcelona can help.