A municipality's citizen service line has been recording every call for four years. No one decided how long to keep the recordings, so they were never deleted. The server holds thousands of hours of audio with names, addresses, phone numbers, complaints about neighbors and, now and then, someone calling in tears because they do not know where else to turn.
One day someone asks what those recordings are for. The honest answer is almost nothing: the reports were already captured, the cases were already closed and no one has listened to them again. But they are still there. Anyone with access to the server can listen to them, copy them or lose them.
The scene repeats under other names. Forms that ask for a national ID number (CURP) to report a broken streetlight. Spreadsheets with data on detained people that circulate by email between departments. Photos of ID cards stored on an operator's personal phone. Each copy was born from a real need or from a "just in case". None of them has an exit date.
In public operations, keeping more is often assumed to be the prudent choice. Mexican law says the opposite. Available data is not necessary data, and data that is no longer necessary is not an asset: it is a risk the institution keeps carrying while getting nothing in return.
On March 20, 2025, Mexico's Official Gazette (DOF) published a new General Law on the Protection of Personal Data Held by Obligated Entities (LGPDPPSO). It applies to any authority, entity, body or agency at all three levels of government, including municipalities. It changed who oversees compliance: the role that the INAI used to play now belongs to the Ministry of Anti-Corruption and Good Governance and to the oversight authorities that the law itself defines, depending on each institution's sphere. What did not change, and is worth reading carefully, are the obligations on how much is collected and how much is kept.
Article 10 lists the principles every data controller must observe: lawfulness, purpose, loyalty, consent, quality, proportionality, information and accountability. Three articles bring them down to operations:
- Article 19, proportionality. The controller "shall only process personal data that is adequate, relevant and strictly necessary for the purpose that justifies its processing". The key word is strictly. It is not enough that a piece of data might be useful someday.
- Article 17, quality. When data is no longer necessary for the purpose that motivated its processing, it must be deleted, after blocking where applicable, once its retention period ends. And those periods "shall not exceed those that are necessary", taking into account administrative, accounting, tax, legal and historical considerations.
- Article 18, procedures. The controller must establish and document how it retains, blocks and deletes data, with defined retention periods, mechanisms to meet those deadlines and a periodic review of whether the data still needs to be kept.
The law adds two pieces that connect all of this to systems. Article 24 requires software platforms and applications that process personal data to comply "by default" with the law's obligations. And Article 27 requires policies that consider the full data lifecycle, "its collection, use and subsequent deletion", along with an inventory of personal data and of the systems that process it.
Put another way: for every piece of personal data that enters operations, the institution should be able to say why it asks for it, who uses it, how long it keeps it and what happens when that time is up.
Excess data almost never comes from a decision. It comes from reasonable habits that no one reviews. In municipal operations, there are four places where it builds up most often.
A citizen report form or a booking form at the municipal holding facility is designed once and used for years. Over time, fields get added because some department needed them for a report that no longer exists, or because "it has always been asked". The result is asking for a CURP, a date of birth and an email address to report a pothole, when all that is needed to fix it is what happened, where, and how to notify the person who reported it.
Every field that does not serve a concrete purpose is data the institution will have to protect, retain, hand over in an access request and, eventually, delete. Removing it from the form is the cheapest way to comply.
Some data is raw material: it is used to produce another record and then becomes surplus. A call recording is used to capture the report. A photo of an ID is used to verify the person making a request. Once the record exists, the question is whether the original input still has a purpose of its own. Sometimes it does, because an applicable rule requires keeping it. Often it does not, and it is kept only because deleting it requires someone to decide.
Spreadsheet exports to put together a report, files emailed to another department, screenshots in messaging groups, folders on the desktop of a shared computer. Every copy outside the main system escapes its permissions, its audit log and its retention periods. If the system deletes a record when its period expires but five copies remain in email, the data was not deleted.
Minimizing is not only about keeping less. It is also about fewer people seeing each piece of data. Article 19 speaks of processing, and looking up a record is processing it. A generic user account shared by ten people, or a profile that sees every case file in every department, multiplies processing with no purpose to justify it.
A word of caution here. In an earlier post we explained why a backup is not an archive and why public information must survive a change of administration. Minimization does not contradict that idea. It completes it.
The law itself defines deletion as an "archival disposal" of personal data under the applicable records rules. In other words, data is not deleted whenever someone feels like it, but when the period set by the institution expires, and that period is coordinated with the General Archives Law. That law requires every obligated entity to have a records disposition schedule: the register that sets out the documentary value, validity, retention periods and final destination of each series of records.
Between the moment data stops serving its purpose and its final deletion there is also blocking. The law defines it as identifying and retaining data once its purpose has been fulfilled, for the sole purpose of determining possible liabilities, until the statute of limitations runs out. During that period the data cannot be processed. The case file of a detained person, for example, may be needed to respond to a human rights complaint years later. Blocking it makes it possible to keep it for that purpose without leaving it open to any lookup.
The difference, then, is not between keeping and deleting. It is between deciding and not deciding. A mature institution knows what it keeps, why, for how long and under what restrictions. An immature one keeps everything, everywhere, for everyone, with no end date.
A practical way to start is to take one process, such as receiving citizen reports or booking a detained person, and ask four questions about every piece of data captured:
- Why is it requested? If the answer is "just in case" or "it has always been asked", the field is a candidate for removal. If the purpose exists, it should be written down and match the privacy notice.
- Who needs to see it? The department handling the case is not the same as the department producing statistics. The latter almost never needs names.
- How long is it useful? Each type of data should have a retention period tied to a records series, not an "indefinite" by omission.
- What happens when it expires? Blocking, deletion or transfer to the historical archive, with evidence that it happened.
The exercise produces, almost without trying, two documents the law requires: the inventory of personal data and processing systems, and the procedures for retention, blocking and deletion. It also tends to reveal that the biggest risks are not in the main system, but in the copies and in the access rights no one reviewed.
This work has an added benefit when the institution uses artificial intelligence. A model that reads case files, transcribes calls or answers citizens processes personal data just as a person does. The less unnecessary data there is in the system, the less an assistant can expose by mistake. Governing what is kept is also governing what AI can see.
Article 24 requires platforms to comply by default. In practice, that means some minimization decisions need to be built into the system's design rather than depend on the discipline of each operator.
In Agora, Intello's citizen service platform, each call recording is deleted once it is transcribed: the report stays in the ticket and the audio does not pile up on a server. Property tax balance lookups by cadastral key are done with the Treasury's authorization and never show the owner's data. Each municipality operates separately from the others, and each department sees only what concerns it. Residents do not have to create an account to check their tickets: they open a signed link or sign in with their phone number and a code sent by WhatsApp.
In Tribuna, the platform for public safety and municipal civic justice, search results are filtered by the role of the person searching, and semantic search checks permissions on every result. Opening a sensitive case file requires a stated reason, and the audit log records every read and every change, just as it records every biometric comparison. Duplicate detection by CURP, date of birth, address and phone number helps keep a single identity per person instead of several copies of the same record. And implementation includes an institutional configuration stage where roles, permissions and forms are defined, which is the natural moment to ask the four questions about every field.
No platform decides for the institution how long to keep a case file or what its final destination is. That decision belongs to its records coordination office and its transparency unit, according to their legal mandates. What a well-designed system can do is make sure that unnecessary data is not collected, that sensitive data is not viewed without a reason and that every access leaves a trace.
For years, the conversation about data in government revolved around having more: more records, more sources, more dashboards. That conversation is still valid, because an institution without information decides blindly. But the 2025 law adds the other half: having what is necessary, for as long as necessary, within reach of those who need it.
The question for leadership teams is not how much data they have. It is how much of that data they could justify today, one record at a time, to the person it belongs to.
If your institution wants to put order into what it collects, who looks it up and how every access is recorded, see how Tribuna controls access to public safety and civic justice case files, how Agora serves residents without accumulating what is not needed, or request a demo.