Facial recognition technologies
Rights at stake
- Violations of criminal proceeding principles
- Principle of adversarial proceeding – art. 111.4 Cost
- Atypical evidence – the (in)applicability of art. 189 CPP
- Uncertain assessment by the judge – art. 192 CPP and the need of corroboration evidence
- Principle of equality of arms – art. 111.2 Cost
- Rights of third parties
Sources of law
- International level – art. 8 ECHR
- EU sources – GDPR, LED, EU Ethical Charter
- Italian framework – SARI
- Decreto Capienze
- “Gates” provided by the actual code
- The revolution of the AI Act
- Forbidden practices and unacceptable risk: art. 5
- Safeguards provided
- High-risk AI systems (art. 6)
- Safeguards for ex-post FRT
- AI Act pros and cons
- Delegation Law 132/2025
Facial recognition technologies are AI tools capable of capturing biometric data (defined in GDPR: those data that can allow who processes it to unequivocally recognize the person) and transforming them into an algorithmic string. Nowadays they are used for everything, even for unlocking our phones and, in some cases, even in the criminal justice system.
These instruments compare, in the span of seconds, those faces with others contained in specific databases, for different objectives.
According to the level of intrusiveness and to their purpose, we find there is a four stages process:
- Gathering: detention of data in databases
- Grouping similar faces in the same group
- Identification: one to many comparison
- Authentication: one to many comparison
- Classification: pulling out information from the picture.
There are two types of FRT:
- Ex-post recognition, which involves static images (es. picture, frame of a video...). The databases used for this kind of comparison contain millions of algorithmic strings, and its main goal is identification (to attribute a name to a face). Useful, for example, to find out who was at the crime scene at the time of the crime.
- Live recognition. These kinds of systems capture people’s faces with no distinction, using live cameras installed in crowded places. In a few seconds, it's capable of collecting a huge quantity of data.
- The database used for this comparison is smaller and groups people who have common criminal records --> for this reason, the database is often called “blacklist”. The main goal here is authentication (useful to find wanted people), with the general purpose of preventing crimes.
- Es. We are looking for a person, and we want to know if that person is at the stadium.
Rights at stake
- Right to have a personal identity --> The Italian Constitutional Court recognizes the right of any person to be recognized for who she really is. These instruments have a high error rate, which clashes with this right.
- Right to privacy and data protection (ovvio, they can see where you are anytime you go to a specific place, parade, manifestation)
- Freedom of association and expression
- Right to self-determination
The last two can be grouped, as they are both united by the Chilling effect (when individuals or groups of individuals avoid to express a right, as they fear they will face criminal consequences): if a person is aware that in a specific place there is a camera, maybe she will avoid going to that place to not be “tracked”. Especially from places where it is possible to pull out “sensitive data” (es. Church, political parade).
Violations of criminal proceeding principles
The use of FRT implies the violation of some traditional principles inside criminal proceedings, such as:
- Adversarial proceeding (art. 111.4 Cost)
- Atypical evidence (art. 189 CPP)
- Uncertain assessment by the judge (art. 192 CPP) and the need of corroboration of evidence
- Equality of arms (art. 111.2 Cost)
- Rights of third parties
Principle of adversarial proceeding – art. 111.4 Cost
According to this well-established principle (referred to by the Italian doctrine as strong adversarial proceeding), the defendant must be able to participate in the formation of evidence.
This means that the defense shall have the chance to debate and respond to all charges. Es. In case of witness testimony, the defendant must be in the position to interrogate him.
The process by which the device reaches the percentage of match is secret; the defense lawyer has no way to counter the reliability of the device when it is inserted as evidence.
Atypical evidence – the (in)applicability of art. 189 CPP
“Quando è richiesta una prova non disciplinata dalla legge, il giudice può assumerla se essa risulta idonea ad assicurare l'accertamento dei fatti [187] e non pregiudica la libertà morale della persona [642, 188]. Il giudice provvede all'ammissione, sentite le parti sulle modalità di assunzione della prova.”
As FRT is not explicitly listed in our criminal procedure code, the only way it can be entered into evidence is by making it atypical evidence, regulated by article 189 CPP.
The problem is that US research (2019-2020) demonstrated that these instruments are almost always correct when talking about white Caucasian males, but there are many other categories of people (woman, people with glasses, black men etc.) for which the percentage of mistake is higher. The machine output depends on information and data provided by the creators, and by self-learning; therefore, the conclusions given by the AI tool may be biased.
According to our code, atypical evidence can be admitted by the judge if adequate to << ensure the fact-finding >>. FRT is digital evidence, included under the more general concept of “scientific evidence” which implies the reach of some requirements to be admitted. The jurisprudence has “codified” some rules for the method to be adopted to the novel science evidence. The judge is basically charged for being the guardian of the proceeding, as he must not allow the entrance of junk science. To ascertain that, there is the so-called “Daubert test”: an obligation upon the judge to issue a reinforced motivation: he must activate a prior adversarial debate to verify the quality of the science introduced in the proceeding. Th
-
Appunti di Industrial Technologies
-
Schemi Operations management and digital technologies
-
Dimostrazioni Communication Technologies
-
Industrial Technologies