-

Key risks include privacy intrusion, biometric surveillance, and bias or discrimination due to uneven model performance across demographics. Legally, facial images may qualify as biometric data, triggering GDPR Article 9 protections, strict consent requirements, and heightened obligations around security, transparency, and purpose limitation.

Bias & discrimination across demographics • Biometric data → GDPR Article 9 → explicit consent required • Data security, privacy, transparency • Automated decisions → Article 22 implications

Yes deepfakes can be haramful

Using facial images raises privacy and consent concerns because biometric data is sensitive. There is also a risk of misuse or biased performance across different demographic groups. Legally, the system must ensure secure storage, transparency, and clear limits on how images are used.

Privacy risk: Collection, storage, and processing of facial images can violate data protection laws (e.g., GDPR, CCPA) if not handled properly. Consent issues: Users must give informed, explicit consent for biometric data use. Data security:

Biometric data processing risk and transparency and user autonomy

Constant camera monitoring can lead to a "chilling effect" where drivers feel micromanaged, causing stress and reducing job satisfaction. There is also the risk of "function creep"—where data collected for safety is later used to judge performance or determine insurance premiums

Peoples facial images cannot be shared publicly. Drivers must know that their faces are being assessed for drowsiness

The use of facial imagery for drowsiness detection raises significant ethical and legal risks, primarily due to the intrusive nature of facial data and its potential misuse. From an ethical perspective, facial images are deeply personal and closely tied to identity. Continuous monitoring may create feelings of surveillance, loss of autonomy, and reduced trust, particularly if drivers are unclear about how long data is stored or how it may be reused. Bias in facial landmark recognition models further exacerbates ethical concerns, as inaccurate detection for certain racial or physical characteristics may disproportionately affect specific groups, reinforcing inequality. Legally, facial imagery constitutes biometric data when processed to uniquely identify or analyse individuals. This creates heightened obligations under GDPR, including strict conditions for lawful processing, transparency, and security. Any failure to clearly define purpose, limit retention, or protect the data could expose IRIS operators to regulatory enforcement and liability.

Privacy, ethical concerns, false negatives. Inaccuracy

Privacy and consent, misinterpretation of facial data

The application taking user biometric data

Privacy & consent: Facial images are biometric data, requiring explicit GDPR consent. Surveillance concerns: Continuous monitoring may be seen as intrusive. Liability: Misclassification causing accidents may expose providers to legal claims.

Respect of persons. We still have several issues when it comes to facial detection with AI because of dataset most models have been trained with.

Some faces may appear drowsy, implication of illegal emotion detection

Risks like the safety of user data collected and the potential infringement on their privacy.

The ethical risks is that the system must be able to give permission before automatically accepting it. A legal risk is the system being able to identify the person using personal details. I think the system should only be used to recognise drowsiness and other facial images such as when the user has fainted. It shouldn't store personal information to be used to identify the person and know who it is.

Collection of personal data and obtaining consent for use. Basis on which such data can be used in legal processes. Recognition of flaws in technology relating to age,ethnicity, gender etc and how that would be embedded.

privacy & data protection; bias and fairness (different skin tones, age groups, facial characteristics, disabilities, facial paralysis); reliability and safety; accessibility and inclusivity (prescription glasses, sunglasses, face masks, facial hair, head coverings); transparency and accountability; legal compliance.

Because the technology has the potential to prevent serious road accidents and save lives, many users may view the safety benefits as outweighing concerns around privacy or data collection. However, this creates its own ethical challenge: the perceived public benefit should not justify overlooking the needs of minority groups or accepting systems that perform unequally across different populations.

Primary risks are privacy, bias across demographics, and overreliance on a system that can still miss genuine drowsiness. Any deployment should minimize personal data and document how false negatives are handled.

Facial-image monitoring risks unfair errors for people with different skin tones, facial features, masks, glasses, or head coverings, plus privacy concerns. A deployment should be narrowly scoped to drowsiness detection only.

The main risks are discrimination, false negatives, privacy intrusion, and unclear liability if the system fails. A safe deployment needs strong performance evidence across varied groups.

Facial monitoring creates risks around biometric privacy and unequal accuracy across demographic groups. It also raises issues if the system is used beyond drowsiness detection.

The biggest risks are unfair treatment, privacy harms, and wrongful alerts that could affect drivers’ confidence or jobs. Systems like this should be evaluated for differential performance.

The ethical and legal risks include unfair false alarms, missed fatigue events, and collecting more data than is necessary. The system should be designed with privacy by default.

The main concern is that an apparently objective safety tool may still encode bias and create legal exposure if decisions are unfair. That is especially true for biometric systems.

The main risks are consent, model bias, and the possibility of identity misuse. A system designed for fatigue should not become a general surveillance tool.

Ethical and legal risks include unfair false alarms, missed fatigue events, and collecting more data than is necessary. The system should be designed with privacy by default.

The key risks are privacy invasion, bias, safety failures, and unclear responsibility when the system gets it wrong. A fair system must be reliable across different users and conditions.

• Identification of illegal drivers. • Identification of criminals – unintended. • Managing expectations of different stakeholders

Nature of the data, how it is stored and how it is used. Whether it is fully anonymized, consent.

Constant monitoring - almost as if you are under constant surveillance whilst driving which could cause people to become more distracted whilst driving as they will be paying additional attention to what their face is doing and how the facial imaging system maybe perceiving their expressions. Over reliance - may result in people driving for longer periods of time than they usually would all because the facial imaging system has not yet informed them that they are tired - they may not look tired but could feel it. Inaccuracy of facial imaging detection.

Collected facial images storage leak! Illegal use. Privacy. Who owns it.

Privacy issues. Is there an explicit consent form for the driver to complete? Does the camera angle only capture the driver or other passengers? How do you validate the system’s accuracy in predicting user behaviour?