Facial Recognition, Emotional AI, and the EU’s Proposed AI Regulation

Facial recognition technology has become widespread in many parts of the world, particularly in Asia. Closely linked to it is a more controversial technology: emotion recognition. This involves AI systems attempting to detect emotional states through facial expressions, eye movements, tone of voice, or other behavioural cues, then drawing conclusions or making decisions based on that data.

In this article, Maria Stefania Cataleta examines the European Commission’s 2021 draft Artificial Intelligence Act and argues that it represented a major legal development, especially regarding surveillance technologies and facial recognition.

Emotional Recognition and Human Rights Risks

The author argues that emotion recognition systems often fail to account for:

• cultural differences

• personal variation

• contextual nuance

• disability-related expression differences

As a result, such systems may reinforce stereotypes, discrimination, and social conformity.

She contends that until robust safeguards exist, emotion-recognition AI should be heavily restricted, especially in public spaces.

Recommended protections include:

• prior notice to individuals

• meaningful consent

• right to opt out

• human oversight

• AI assisting, not replacing, human judgment

Surveillance and Democratic Concerns

The article places facial and emotional recognition within a broader concern about mass surveillance.

According to the author, unchecked use of AI by authoritarian states could threaten:

• privacy

• freedom of thought

• free expression

• movement

• autonomy

• democracy itself

She warns that surveillance cities and algorithmic political conditioning could normalize social control if left unregulated.

The EU’s Proposed AI Regulation

The 2021 European Commission proposal aimed to create a harmonized AI legal framework across all European Union member states.

The draft was presented as a successor to the General Data Protection Regulation, with even stronger penalties—up to 6% of global annual turnover, compared with GDPR’s 4%.

Its goals included:

• safe and trustworthy AI

• innovation-friendly regulation

• protection of fundamental rights

• common standards across Europe

Practices Proposed for Prohibition

The article highlights several banned or highly restricted uses of AI:

1. Subliminal Manipulation

AI systems using techniques beyond a person’s awareness that could cause harm.

2. Exploitation of Vulnerable Groups

Systems targeting children, elderly persons, or people with disabilities to distort behaviour.

3. Social Scoring

Public-authority systems rating people based on behaviour or personal traits, especially where outcomes are unfair or disproportionate.

4. Real-Time Remote Biometric Identification in Public

Live facial recognition in public spaces for law-enforcement purposes, except in narrow security circumstances and subject to prior authorization.

Transparency Requirements

Certain AI systems such as chatbots would need to disclose that users are interacting with AI, allowing informed decisions about whether to continue.

Low-risk systems would face lighter obligations.

Relationship with GDPR

Cataleta notes similarities with GDPR principles, especially around:

• accuracy

• security

• robustness

• resilience

• accountability

The AI framework was therefore positioned as an evolution of Europe’s earlier data-protection model.

Broader Global Argument

The article concludes that Europe was attempting to become a global leader in rights-based AI governance.

The author also argues that international regulation should eventually move beyond declarations toward a binding treaty on AI, though she acknowledges this would be difficult in the current geopolitical climate.

Core Takeaway

Cataleta’s central claim is that AI should serve individuals rather than control them. Europe’s regulatory model sought to combine innovation with safeguards so that technologies like facial recognition and emotion recognition remain under human control and compatible with democratic values.