---
title: "EU AI Act for Insurance AI Voice Agents (2026 Guide)"
description: "Claims and renewal calls by an AI agent need an AI disclosure from August 2026. Life and health pricing is high-risk from December 2027. Where the line is."
url: "https://ainora.lt/blog/eu-ai-act-insurance-ai-voice-agents-2026"
published: "2026-11-11"
author: "Justas Butkus"
---

# EU AI Act for Insurance AI Voice Agents (2026 Guide)

> **TL;DR:** For most insurance phone work - claims intake, renewal reminders, quote follow-ups and policy questions - an AI voice agent is not a high-risk system under the EU AI Act. What applies is Article 50: from 2 August 2026 the caller must be told they are talking to an AI, at the latest at the start of the call. The line is crossed when the AI is used for risk assessment or pricing of people in life or health insurance (Annex III point 5(c)), or to judge creditworthiness (point 5(b)). Those uses are high-risk from 2 December 2027 and also need a fundamental rights impact assessment before first use.

> **Not legal advice**
>
> This is general information about Regulation (EU) 2024/1689 as amended by Regulation (EU) 2026/1744. Every legal point links to the article it comes from. All sources were last checked on 11 November 2026. Your own counsel decides how the rules apply to your business.

- **2 Aug 2026** - AI disclosure duty (Article 50) applies (Source: AI Act, Art. 50 - https://artificialintelligenceact.eu/article/50/)
- **2 Dec 2027** - Annex III high-risk duties apply, after the 2026 delay (Source: Reg. (EU) 2026/1744 - https://eur-lex.europa.eu/eli/reg/2026/1744/oj)
- **5(c)** - Annex III point: life and health insurance risk assessment and pricing (Source: AI Act, Annex III - https://artificialintelligenceact.eu/annex/3/)
- **EUR 15M / 3%** - maximum fine for deployer or transparency breaches; SMEs pay the lower of the two (Source: AI Act, Art. 99 - https://artificialintelligenceact.eu/article/99/)

## Is an AI phone agent for insurance high-risk under the EU AI Act?

Usually not. The AI Act does not treat "insurance" as a high-risk industry. It lists specific uses. For insurance, the list in [Annex III](https://artificialintelligenceact.eu/annex/3/) names two: AI used to evaluate the creditworthiness of people or set their credit score (point 5(b)), and AI used "for risk assessment and pricing in relation to natural persons in the case of life and health insurance" (point 5(c)).

An AI agent that answers the phone, takes down a claim, reminds a customer about a renewal or books a call with a broker does none of those things. It talks to people, so it falls under the transparency duty in [Article 50](https://artificialintelligenceact.eu/article/50/): people must be told they are interacting with an AI system, unless that is obvious. On a phone call it is not obvious, so the agent has to say it.

What decides the risk level is the job the system does, not the label on your company. The same agent can be low-risk in one flow and high-risk in another, depending on what its answers are used for.

## Which insurance call uses are high-risk, and which are not?

Here are the AI uses insurers, brokers and agencies ask about most, with the risk level each one gets under the Act as it stands after the July 2026 amendment.

| AI use on the phone | Risk level under the AI Act | What it triggers | From |
| --- | --- | --- | --- |
| Claims intake (first notice of loss): takes the facts, opens the claim, hands over | Not high-risk | Tell the caller it is an AI (Art. 50) | 2 Aug 2026 |
| Renewal and payment reminders | Not high-risk | AI disclosure (Art. 50) | 2 Aug 2026 |
| Quote follow-up, booking a call with a broker | Not high-risk | AI disclosure (Art. 50) | 2 Aug 2026 |
| Policy questions: cover, documents, claim status | Not high-risk | AI disclosure (Art. 50) | 2 Aug 2026 |
| Collecting answers that feed risk assessment or pricing of a person for life or health cover | Can be high-risk (Annex III 5(c)); check with counsel | Deployer duties (Art. 26), impact assessment (Art. 27), disclosure | 2 Dec 2027 (disclosure from 2 Aug 2026) |
| Checking a person’s creditworthiness or credit score, for example for instalment plans | High-risk (Annex III 5(b)) | Deployer duties (Art. 26), impact assessment (Art. 27), disclosure | 2 Dec 2027 (disclosure from 2 Aug 2026) |
| Pricing property, motor or travel cover | Not on the Annex III list | Disclosure if it talks to people; GDPR applies | 2 Aug 2026 for disclosure |
| Recognizing callers’ emotions from their voice | High-risk (Annex III 1(c)) | Tell the people exposed to it (Art. 50(3)), deployer duties | 2 Aug 2026 notice, 2 Dec 2027 duties |

[Source: AI Act, Annex III and Art. 50](https://artificialintelligenceact.eu/annex/3/)

Two details matter in that table. First, point 5(c) covers **life and health** insurance only. Pricing a car or a house is not on the list. Second, the creditworthiness point excludes AI used to detect financial fraud, but that exception belongs to point 5(b) only.

## Why is life and health insurance pricing high-risk?

Because the result decides whether a person can get cover and what they pay for it, and the EU lawmakers put that next to credit scoring as access to an essential private service. The wording is narrow: "AI systems intended to be used for risk assessment and pricing in relation to natural persons in the case of life and health insurance."

Where does a phone agent come in? If it only books a medical underwriting call, it is outside the point. If it asks the health questionnaire and its answers flow into the rating that sets a person's premium, it can become part of a system used for risk assessment and pricing. That is the case to look at with your counsel.

[Article 6(3)](https://artificialintelligenceact.eu/article/6/) gives an exit for Annex III systems that do a narrow procedural task, improve a finished human task, spot patterns without replacing human judgment, or do preparatory work, as long as they do not materially influence the decision. But the same article says an Annex III system is **always** high-risk when it profiles people. Using health answers to predict a person's risk is very likely profiling, so do not count on that exit for underwriting.

## What must an insurance AI agent say at the start of a call?

Article 50(1) requires AI systems that talk to people to be designed so that people are told they are dealing with an AI. Article 50(5) adds that this must be clear and given at the latest at the first interaction. For a phone agent, that means the opening line.

A workable opening for a claims line: "Hello, you are speaking with the virtual assistant of [Agency], an AI. I can take the details of your claim now, or put you through to a colleague." It says what the caller is talking to, what the agent can do, and how to reach a person.

- **Say it first,** before any personal questions.
- **Say it plainly.** "Virtual assistant" alone can be vague; adding "an AI" removes the doubt.
- **Repeat it when asked.** If a caller asks "am I talking to a real person?", the answer is always a clear no.
- **Outbound calls too.** Renewal and payment reminder calls need the same opening.

If the agent analyzes callers' emotions from their voice, for example to flag stress, [Article 50(3)](https://artificialintelligenceact.eu/article/50/) adds a second notice: the people exposed must be told that emotion recognition is running.

## If your use is high-risk, what must the insurer do?

An insurer that uses a high-risk system built by someone else is a "deployer". The deployer duties are in [Article 26](https://artificialintelligenceact.eu/article/26/), with one extra duty in Article 27 that applies to life and health pricing and credit scoring by name. They apply from 2 December 2027.

1. **Use it as the provider instructs** - Take technical and organizational measures so the system is used in line with its instructions for use. (Source: AI Act, Art. 26(1) - https://artificialintelligenceact.eu/article/26/)
2. **Put a competent person in charge of oversight** - Assign human oversight to people with the necessary competence, training and authority, and give them the support to do it. (Source: AI Act, Art. 26(2) - https://artificialintelligenceact.eu/article/26/)
3. **Keep the logs** - Keep the logs the system generates, under your control, for at least six months, unless other EU or national law says otherwise. (Source: AI Act, Art. 26(6) - https://artificialintelligenceact.eu/article/26/)
4. **Tell the people affected** - Where the system makes or helps make decisions about people, inform those people that a high-risk AI system is used. (Source: AI Act, Art. 26(11) - https://artificialintelligenceact.eu/article/26/)
5. **Do a fundamental rights impact assessment first** - Deployers of systems under Annex III points 5(b) and 5(c) must assess the impact on fundamental rights before first use: the process, the period and frequency of use, the people affected, the risks and the oversight measures. (Source: AI Act, Art. 27 - https://artificialintelligenceact.eu/article/27/)
6. **Be ready to explain decisions** - A person affected by a decision based on the system’s output, where it has legal effects or similarly significant effects on them, can ask for clear and meaningful explanations of the role the AI played and the main elements of the decision. (Source: AI Act, Art. 86 - https://artificialintelligenceact.eu/article/86/)

If you build the system yourself, you are also its provider, and the provider duties in [Article 16](https://artificialintelligenceact.eu/article/16/) are yours too. Breaching the deployer duties or the transparency duty can cost up to EUR 15 million or 3% of worldwide annual turnover, whichever is higher; for SMEs and start-ups the lower of the two applies ([Article 99](https://artificialintelligenceact.eu/article/99/)).

## What are the key dates for insurers and brokers?

Regulation (EU) 2026/1744, published on 24 July 2026, moved the high-risk dates. The AI disclosure date (Article 50(1)) did not move.

| Date | What applies | What it means for insurance calls |
| --- | --- | --- |
| 2 Feb 2025 | Prohibited practices (Art. 5) and AI literacy (Art. 4) | Staff who run AI tools need support to understand them; manipulative or exploitative AI is banned |
| 2 Aug 2026 | Transparency duties (Art. 50) | Every AI phone agent tells callers it is an AI |
| 2 Dec 2027 | Annex III high-risk duties, after the 2026 delay | Life and health pricing and credit scoring uses follow Art. 26 and Art. 27 |

[Source: Reg. (EU) 2026/1744](https://eur-lex.europa.eu/eli/reg/2026/1744/oj)

The AI literacy duty was softened in July 2026: providers and deployers must "take measures to support the development of AI literacy" of their staff, and the article now says it does not require any specific level for any individual ([Article 4](https://artificialintelligenceact.eu/article/4/)).

## How do you keep an insurance AI agent on the simple side of the line?

Most insurers want the phone agent for speed and availability, not for underwriting. If that is your goal, a few design choices keep it a transparency-only system and make the review with your counsel short.

- **Keep it out of the price.** The agent records what the caller says and passes it on. The premium comes from your underwriters or your existing rating process, which you assess on its own.
- **No health questionnaire by phone AI** unless your counsel has approved that flow and its classification.
- **Disclosure in the first sentence,** on inbound and outbound calls.
- **Clear handover rules:** complaints, distressed callers, disputes about a claim decision and anyone who asks for a person go to a human.
- **Write down your classification.** One page: what the agent does, what it does not do, why it is not in Annex III. If the flow changes, update it.

## Does GDPR still apply on top of the AI Act?

Yes. The AI Act adds rules for the AI system; GDPR still governs the personal data in every call. Insurance calls often contain health details, which are a special category of data under GDPR, so access to recordings and transcripts should be limited to the people who need them. GDPR's rules on decisions based solely on automated processing also stay in force alongside the AI Act. See our [GDPR guide for AI voice agents](/blog/ai-voice-agent-gdpr-compliance-guide) for the data side.

## How does an AI phone agent from AInora fit?

AInora builds AI phone agents that can answer and make calls for insurers, brokers and agencies: claims intake, renewal reminders, quote follow-ups and policy questions. How it lines up with the points above:

- The agent says it is an AI assistant in its first sentence, in the words you choose.
- It hands the call to a person when the caller asks, or when your rules say so.
- Every call gets a transcript and a summary your team can review.
- You decide what it may and may not ask. If health questions are off limits, they stay off limits.
- It speaks many languages, so the same rules hold for every caller.

Your counsel or DPO decides the classification and the rules; the agent follows them. To hear it on your own use case, [fill out the short form](/contact?from=blog-eu-ai-act-insurance-ai-voice-agents-2026) and we build a demo agent from your website. More on the use cases: [AI voice agents for insurance agencies](/blog/ai-voice-agent-insurance-agencies-2026).

## Frequently Asked Questions

**Is an AI receptionist for an insurance agency high-risk under the EU AI Act?**

Not when it answers questions, takes claim details, books calls or sends reminders. Annex III lists only life and health insurance risk assessment and pricing (point 5(c)) and creditworthiness checks (point 5(b)) for this area. An AI receptionist must still tell callers it is an AI under Article 50 from 2 August 2026. (Source: AI Act, Annex III - https://artificialintelligenceact.eu/annex/3/)

**Is motor or property insurance pricing with AI high-risk?**

Not under Annex III. Point 5(c) covers risk assessment and pricing of natural persons in life and health insurance only. GDPR and the transparency duty still apply. (Source: AI Act, Annex III 5(c) - https://artificialintelligenceact.eu/annex/3/)

**When do the high-risk rules for insurance start?**

They apply from 2 December 2027. The original date was 2 August 2026; Regulation (EU) 2026/1744 moved it. The Article 50 disclosure duty was not moved and applies from 2 August 2026. (Source: Reg. (EU) 2026/1744 - https://eur-lex.europa.eu/eli/reg/2026/1744/oj)

**Does an insurer need a fundamental rights impact assessment?**

Article 27 requires it from public bodies, private entities providing public services, and deployers of systems under Annex III points 5(b) and 5(c). So an insurer using AI for life or health pricing, or for credit scoring, must do one before first use. A claims intake agent does not trigger it. (Source: AI Act, Art. 27 - https://artificialintelligenceact.eu/article/27/)

**Can an AI agent ask health questions on the phone?**

The AI Act does not ban it, but if the answers feed life or health pricing, the system can fall under Annex III point 5(c), and the health data is a special category under GDPR. Decide that flow with your counsel before switching it on.

**Can the agent detect stress in a caller’s voice to flag fraud?**

Emotion recognition on customers is not banned, but it is listed as high-risk in Annex III point 1(c), and the people exposed must be told about it under Article 50(3). The workplace ban in Article 5(1)(f) covers employees, not customers. (Source: AI Act, Art. 50(3) - https://artificialintelligenceact.eu/article/50/)

**What is the fine for not disclosing an AI agent?**

Breaches of the Article 50 transparency duties can be fined up to EUR 15 million or 3% of worldwide annual turnover, whichever is higher. For SMEs and start-ups the lower of the two applies. (Source: AI Act, Art. 99 - https://artificialintelligenceact.eu/article/99/)

Related on AInora

Explore the platform and industry pages relevant to this article.

- [AI for insuranceClaims intake and policy questions](/industries/insurance)
- [AINORA AI voice agentPlatform overview and capabilities](/ai-voice-agent)
- [AI debt collectionCompliant voice for recoveries](/ai-debt-collection)
- [PricingPriced individually, after a call](/pricing)
- [How it worksSetup, integrations, and go-live](/how-it-works)