Emotion Recognition and EU AI Act Article 50: Does Your Sentiment Tool Need a Disclosure?

Most businesses know Article 50 requires a disclosure when a chatbot talks to customers. Fewer realise that Article 50 also contains a separate disclosure obligation — Article 50(3) — covering any AI system that analyses people's emotions, infers their psychological state, or categorises them biometrically.

If your customer service platform analyses caller sentiment, if your support chat scores customer frustration in real time, if your website uses engagement tracking that infers emotional state — this obligation may apply to you from 2 August 2026.

What Article 50(3) actually requires

Article 50(3) of the EU AI Act states that deployers of AI systems used for emotion recognition or biometric categorisation must inform the natural persons exposed to those systems that they are subject to such processing. This is a separate obligation from the chatbot disclosure under Article 50(1).

The key words are "natural persons exposed to" — meaning the people being analysed, not your operators or staff reviewing the outputs. If your system analyses a customer's voice tone to infer frustration, that customer must be informed. If your platform scores chat messages for sentiment and uses that to route conversations or trigger escalations, the customers being scored must be informed.

Which tools are commonly in scope

Customer support sentiment analysis

Many support platforms — including features in Zendesk, Intercom, Freshdesk, and Salesforce Service Cloud — offer sentiment analysis that detects customer frustration, satisfaction, or urgency from chat or voice inputs. If these features are active and applied to your EU customers, the customers being analysed must be informed under Article 50(3).

AI voice agent tone analysis

Some voice AI platforms analyse caller tone, speech patterns, or emotional cues to route calls, adjust script delivery, or flag at-risk interactions. If your voice agent or telephony platform uses this kind of analysis on EU callers, disclosure is required.

Chatbot emotion scoring

Some chatbot platforms score user messages for emotional signals and use this to trigger escalation workflows, adjust response tone, or generate customer health scores. If your chatbot uses emotion scoring on EU users' messages, those users must be informed.

What is probably out of scope

Pure keyword-based routing ("if message contains 'cancel' → route to retention team") without inferring emotional state is unlikely to be covered. Basic CSAT surveys where the customer actively rates their experience are not covered — the customer is providing the rating, not being analysed. Back-office analytics that aggregate sentiment data without exposing the analysis to individual users at interaction time may also be outside scope — but legal review is recommended before concluding this.

What the disclosure must contain

Article 50(3) requires the person to be "informed." The Commission's draft guidelines indicate this means they must know they are subject to emotion recognition or biometric categorisation processing. The disclosure doesn't require revealing the specific outputs or scores — it requires informing the person that this type of analysis is happening.

In practice, this means adding a disclosure notice at the point of interaction where emotion analysis occurs. For a support chat, this might appear alongside (or as part of) the chatbot disclosure: "You're chatting with an AI assistant. This session may use sentiment analysis to help route your query." For a voice call, it's part of the opening script: "This call uses AI to help us respond to you. Conversation analysis may be used to improve your experience."

The important exception: Article 5 prohibition

Before thinking about Article 50(3) disclosure, first check whether your use case is prohibited entirely. Article 5 of the EU AI Act bans certain emotion recognition applications in specific contexts — including the use of emotion recognition in workplaces and educational institutions. If your system falls under Article 5, no disclosure will make it compliant. It must be discontinued.

Article 50(3) covers permitted uses only. If your use case is allowed (customer service, retail, general consumer interaction), then the disclosure obligation applies. If it's prohibited by Article 5, disclosure is not the solution.

What you need before 2 August 2026

1. Audit your tools. Check every customer-facing platform for emotion recognition or sentiment analysis features. Look in: support chat settings, voice platform documentation, CRM AI features, engagement tracking tools. Many of these are turned on by default and may not be obviously labelled as "emotion recognition."
2. Check Article 5 first. For each system, confirm the use case is not prohibited under Article 5 before proceeding. Particularly if the analysis involves employees, students, or vulnerable groups.
3. Add disclosure at the point of interaction. Wherever emotion analysis applies to EU users, add a notice. This can be combined with your Article 50(1) chatbot disclosure if both apply.
4. Log the disclosure. Same requirement as for chatbot disclosures — you need a timestamped record that the disclosure was shown, when, and to whom (at session level, not necessarily individual user level).

The deadline and the fine

Article 50(3) is enforceable from 2 August 2026. No grace period. Fines for non-compliance reach up to €15 million or 3% of global annual turnover, whichever is higher. This is the same penalty tier as the chatbot disclosure violation.

The free Article 50 Scope Check at disclo.eu includes emotion recognition in its scope assessment — if your setup includes sentiment analysis tools, it will flag this obligation and tell you what you need to do.