Skip to content
megapromotingLet's talk
Products sesizari.md

Civic platform Public platform

A problem in the city, described so it can be tracked.

sesizari.md organises reports about public space: location, category, description and relevant materials. The platform focuses on documentation, moderation and data protection before publication.

sesizari.md homepage: the title “You report a problem in public space”, the note that faces and licence plates are covered before publication, and the reporting button
sesizari.md · the anonymisation note is in the hero, not in the footer

sesizari.md

From information to work done.

01

Documentare

The user describes the situation and the location, with the information needed to understand it.

02

Verificare

The materials go through moderation and personal data protection steps.

03

Tracking

The case file and its updates keep the context of the report.

Where it becomes useful.

Public space

Waste, infrastructure, accessibility and other situations organised by category.

Civic documentation

Structured information, easy to review and complete.

Transparency

A public archive of approved cases, with limits on the information published.

A published report does not automatically mean a confirmed violation or a resolved issue. The status of each case is checked in the platform.

sesizari.md in detail

What you can do with this project.

A report only becomes useful when a public official can read it within the legal deadline. sesizari.md takes a photo, a location and a description and turns them into a case file with a completeness score, a category, a competent authority and a petition text — then tracks the 30-day deadline.

The difference from what existed before in Moldova is what is not published. The registration number goes into the internal file to the authority and never appears publicly. The public location is rounded. No photo reaches the site without a human-approved blur. These rules are not policy promises: they are constraints in the database and in the site security header, verifiable by anyone.

01

A case file, not a form

Five categories — illegal parking, abandoned vehicles, waste, unauthorised street advertising, vandalism — with a catalogue of 110 concrete types of acts behind them, each with the legal article indicated as guidance. A case file can contain up to ten objects (for example, ten cars in a yard) and up to 33 photos. Ten is not a random number: beyond that, the file can no longer be read by a person within thirty days.

02

Completeness score, with explanations

Each case file receives a score between 0 and 100 across seven criteria: photos, confirmed location, date and time, a description of at least 40 characters, context, an identifying element and the risk of personal data. Above 90 is „very good”, above 70 „good”, above 50 „poor”, below — „we do not recommend submission”. Missing items are shown as concrete sentences, not as a percentage.

03

Two paths, chosen by the citizen

Anonymous reporting is the default path and remains the default: the platform must not make identification the easy route. Signed petition is the second path — the citizen identifies themselves to the authority, and only then does the 30-day deadline from article 76 of the Administrative Code start. Anonymous petitions are not examined, and the platform says this instead of leaving the opposite illusion.

04

Moderation with a closed gate by default

Nothing is published without human confirmation. The raw photo remains private evidence; only a redacted derivative is made public, with the applied regions, preserved provenance and the moderator's explicit approval — and the person who redacts is not the one who approves. If the media becomes unsafe, the case is withdrawn from public view automatically.

05

The deadline, tracked without cron

The 30 days in Article 60 paragraph (1) of the Administrative Code are calculated at read time, from the sending date, so there is no flag that can lag behind. The moderation queue warns from day 25, so someone can insist before the deadline passes.

06

A signal of trust for the moderator

Each case receives a confidence score from nine signals: identity, account age, reporter history, on-site confirmations, photo evidence, EXIF time, location, duplicate image, and burst submissions. It starts at 50 — “we know nothing”. A low band does not mean rejected, it means “you look first”: the queue is sorted in ascending order by score. Duplicates are caught with a 64-bit perceptual fingerprint, which withstands recompression and light cropping.

07

The appropriate authority, checked by hand

The catalogue of authorities — city hall, sector administrations, police, environmental inspectorate — was completed by copying each email, phone number and address from the institution’s official website. Where an email could not be confirmed on an official page, the field was left empty, with the reason noted. Nothing was filled in from memory or from aggregators.

Data and operation

What enters the system. What needs to be checked.

Personal data do not reach the public — at database level
The registration number, faces, reporter data and exact coordinates never appear in public projections. The last line of defence is in the two free-text fields written by a moderator: publication is refused if the text contains a Moldovan car number, a phone number, an email address or an IDNP. Names are not searched with regular expressions — they cannot be detected correctly; they remain a moderation rule, stated as such.
The public position is generalised
The case has two location columns: one exact, private, which goes to the authority, and one public. The public projection rounds coordinates to three decimal places — a cell of approximately 111 by 76 metres at Chişinău’s latitude. Enough for the issue to be findable, insufficient for a house to be pointed at.
Closed system — nothing leaves the browser to third parties
No resource, request or redirect from the citizen’s browser to a host that is not the site’s own origin: no CDNs, external fonts, third-party map tiles, external geocoding, analytics, external captcha or embeds. The content security policy served by the site states it publicly — `default-src 'self'`, with no host exceptions — and anyone can read it with a single command.
Artificial intelligence is off by default and suggests, not decides
Automatic assistance starts only if the flag is exactly `true` and the key exists; a flag on its own means “off”, not “half on”. The model runs in a deployment in the European Union, has a per-case call cap, and using a model from outside the EU requires a separate flag and a signed document. No automated decision without human confirmation — a requirement of Law 195/2024.
Automatic blurring of faces and number plates: no, for now, and why
A local detector, without an external service, would have been the right path. Almost everything that exists pre-trained for number plates is derived from Ultralytics YOLO and distributed under AGPL-3.0, which would require publishing the entire application under the same licence. The only two candidates with a clean licence are not in the required format and require a custom conversion. We wrote down what we searched for and why we stopped, instead of delivering a detector with a licence in the fog.

From exploration to implementation

How we prepare a project with sesizari.md.

01

We set up the circuit

Which categories fall within scope, who moderates, who approves publication, and which authority each type of offence goes to. Routing relies on an authority catalogue checked against official sources, not on assumptions.

02

We configure the publication gate

What is masked, who confirms, what remains internal. The gate is closed by default: if the wording has not been approved, publication cannot be made — not as a written policy, but as a database restriction.

03

We check the deadlines and responses

We submit a test file, follow the registration with the authority and the official response, and confirm that the day 25 warning and the delay marker appear when they should.

04

We test what must not be seen

Before any launch, the reverse scenario is run: a licence plate intentionally entered into a public field must block publication, and searching for it must return zero results.

Questions worth clarifying.

Does the licence plate in my photo become public?

No. It goes exclusively into the internal file sent to the authority. It does not appear in public, neither in the photograph — public photographs are only blurred, human-approved derivatives — nor in the text: publication is automatically refused if a moderator, out of habit, writes a car number, a phone number, an e-mail or an IDNP in a public field. This is the deliberate difference from the previous civic platform in Moldova, which published them openly.

Does a published report mean that someone broke the law?

No. It means that a situation was documented and, if it was sent, that an authority was notified. Identifying the offender is the authority’s job, not the platform’s, and the legal article indicated next to each type of offence is indicative and must be confirmed against the current text in force. Beneath the 110 types of offences in the catalogue, it says exactly that.

Can I report anonymously?

Yes, and that is the default option — anonymous reporting. But it is fair to know what that means: anonymous petitions are not examined, according to Article 76 of the Administrative Code, so the 30-day deadline does not start running and no official response comes. For that, you need the signed petition route, where you identify yourself to the authority. The platform leaves you the choice and does not push identification as the easy path.

Will it be seen exactly where I live if I report something near my home?

No. The file keeps two locations: the exact, private one that goes to the authority, and the public one, rounded to three decimals — about 111 by 76 metres in Chişinău. The map shows the area, not the yard.

What is loaded into my browser from others?

Nothing. No Google Fonts, no map tiles from third parties, no analytics, no external captcha, no embeds. You can check it yourself: the site’s `content-security-policy` header says `default-src 'self'` and names no other host. It is the project’s number one rule, not a setting someone changes in an afternoon.

Do you use artificial intelligence on my photo?

Automated help exists — it reads the vehicle’s make, colour and condition from the photo, suggests petition text, helps with blurring — but it is off by default, has a per-file call limit, and runs on a deployment in the European Union. A model outside the EU requires a separate flag and a signed document. And, whatever the model says, nothing is published without a human confirming it.

Can it be used by a town hall or an institution?

The workflow, categories, routing to the authority and moderation rules can be adapted. What we do not assume automatically is access to state systems: integration with public registers or with an institution’s petition systems is a separate discussion, with its own legal basis. What we offer today works from the outside, through the institution’s official petition channel.

Has any report reached an authority through the platform?

Not yet. The transmission pipeline — routing, file, signature, sending, retry, log of each attempt — is built and checked, but so far it has only run in the mode that shows exactly what would be sent without sending it. The first real transmission depends on configuring the production e-mail server. We say this because it is the difference between a platform that works and one that has also been used.

Illustrative example

A car abandoned for months in a yard

A use case, without client data or attributed commercial results.

Initial situation

The citizen photographs the car, confirms the location on the map, chooses the category “abandoned vehicle” and writes what can be seen. If there are several cars, they add them to the same file — up to ten.

How it works

The file receives a score from seven criteria, with the specific list of what is missing. The automated assistant, if enabled, reads the make, colour, body type and signs of abandonment from the photo and suggests the administrative paragraph; the citizen can rewrite it. The registration number read from the photo goes into the internal file. The moderator checks, approves the blurring of the photo and publishes; the system routes the file to the competent authority from the catalogue.

Rezultatul

A public file with title, generalised location, blurred photo and trackable status, plus a complete internal file for the authority. After sending, the 30-day term counts itself, with a warning from day 25.

What is required:A factual description and a confirmed location. For an official response within the legal deadline, a signed petition trail is needed — anonymous information is not examined. The legal article indicated next to the category is for guidance.

Collaboration possibilities

sesizari.md, in the context of your organisation.

Reports and public relations

Registration, moderation and allocation flows for reports, with clear responsibilities and protection of personal information before publication.

Private companies

We define a pilot around a real process: users, data, integrations, costs and acceptance criteria. Expansion follows after evaluating the result.

State institutions and companies

We establish the requirements for accessibility, hosting, data protection and interoperability. Any connection to AGE or STISC services requires validation of eligibility, access and approvals.

These are adaptation scenarios, not statements about existing contracts or partnerships. The proposed functions are confirmed within the project scope.

Discuss a pilot

Part of an ecosystem.

What would you want to work better?

Tell us about your process. Together we decide what is worth building, what we can connect, and how we check the result.

Let's talk

sesizari.md — the civic platform, explained · video in Romanian, with subtitles and a transcript in English

Full transcription

Today, we are presenting the sesizari.md platform and analysing a fascinating process, the way a banal photo of an urban problem ends up being transformed, step by step, into a legally structured official dossier. It is a direct look behind the scenes of a system that has a huge impact on the public space around us. So, how does a street photo become an official dossier? When someone notices an irregularity in the city, how do we make that information have a real effect? Well, the system we are dissecting today acts as a bridge. It takes a simple observation, the kind of detail we pass by every day, and turns it into an official report to the authorities. Very important to remember from the start, the platform does this without assigning any blame in advance.

It is simply a clear, objective and very well organised reporting tool. To understand exactly how things work, our analysis will touch on 5 key points. First, what exactly is this platform, then the step-by-step route of a report, the fields and types of problems covered, the strict confidentiality rules and, finally, what that sticker on cars may really mean. Let us start with the first point. What is sesizari.md? And why does it present itself as a neutral tool? The central mission brings an absolutely fundamental distinction. The platform structures reports for the authorities, but, and this is a big but, it does not establish guilt and does not apply sanctions. In practice, we are not dealing with a judge or a reporting officer, the system is a completely neutral bureaucratic intermediary.

It takes raw field data and packs it rigorously, exactly in the legal format required for an official petition. We move on to the route of a report. How does information from the street get directly onto the authorities' desk? Everything unfolds in three extremely clear steps. The first step is the report itself, a photo, marking the location on the map and a short neutral description of the act. Then it goes behind the scenes. Here, personal data are censored, the dossier takes the proper legal form and is sent to the competent institution. The last step is the one that really adds accountability. The system does not just send the email and forget about it, it actively tracks the legal 30-day deadline within which the authority is obliged to respond.

In this way, the whole process becomes traceable and 100% transparent. And a truly fascinating security detail from behind the scenes of this flow is moderation. It is extraordinarily strict. For a photo to become public on the platform, a person, not an algorithm, must manually confirm the redaction, meaning they must check with their own eyes that faces and registration numbers are completely blurred. Without this human validation, publication remains blocked indefinitely. We reach the third chapter. Fields and types of violations. Or, rather, the scope of action. The catalogue of problems that can be reported is truly vast. We are talking about 110 specific types of violations divided into 17 different fields. We can give as examples anything from abandoned cars and potholes in the asphalt to buildings erected without authorisation or illegal street trading.

The most interesting part is that all these 110 categories are active and ready to be taken over right now. They are not just functions left on hold for future updates. But how can such a mountain of diverse problems be managed efficiently? The solution is an intelligent routing system. Parking on a pedestrian crossing or an act of vandalism goes instantly to the police. Litter in green spaces or a car left forgotten for years goes straight to the town hall's desk. And on the other hand, a spill of toxic substances triggers an alert to the environmental inspectorate. This automatic sorting is vital. It guarantees that the information hits the right target from the first second, without getting lost in bureaucratic maze.

We now move on to a critical aspect. Data protection and confidentiality. Here we rely on facts, not promises. Any civic tool automatically raises questions about privacy. The platform addresses this through an impenetrable wall between what is public and what is private. On the public register, visible to everyone, only general data appear, the problem category, the approximate area and the blurred photos. In contrast, the private dossier, the one that goes strictly to the authorities, contains the exact location to the metre, the licence plate numbers and the contact details of the person who submitted the report. These critical details always remain invisible to the general public. There are very clear mechanisms to maintain this safety. Besides blurring faces and plates, the site refuses to use those tracking cookies that usually follow our steps across the internet.

The map uses its own infrastructure, so no external service collects browsing data. The cleverest aspect is identity management. The person reporting is anonymous to the public, but their identity is communicated to the authorities. This happens for a very simple reason. The law requires a real identity, so that a petition can be reviewed and resolved. And so we reach our final point. The meaning of the sticker. What does that sticker we see on the street actually mean? This is probably the most visible and the most misunderstood element of the whole process. If a sticker appears on a vehicle window, it is strictly a physical notification. It simply tells us that a public problem has been reported, and the code on it leads directly to the page of the online report.

The absolutely essential message here is that the sticker is not a fine, it does not establish an offence. Enforcement of the law is, and remains from start to finish, the exclusive responsibility of the competent authorities. Moreover, to keep everything correct and fair, the platform offers the right to contest. Anyone who considers themselves unfairly targeted can request an additional review, during which the dossier is temporarily hidden from the public space. And the physical sticker — the platform rules clearly state that it must be removed immediately upon a simple request. It is an excellent safety net to avoid any kind of abuse. When we draw the line, this whole incredibly well-structured flow works according to a single philosophy.

Facts about how we handle what you give us, not promises. The approach does not come from vague guarantees, but shows concretely and transparently, through solid procedures, that the data is protected, and the complaints reach exactly where they need to for a resolution in line with the law. Which leaves us with a rather thought-provoking final idea. How will this level of transparency change the relationship we have with our cities? When any small observation from the street can be quickly channelled into a process governed by strict deadlines and mandatory responses, we may be witnessing a major shift from silent frustration to measurable and impactful civic action. It is a model that has every chance to reshape the future of urban administration. Until the next analysis, all the best! Goodbye!