Skip to main content
Compliance

EU AI Act: What Your Business Needs to Have in Hand, and When

What does the EU AI Act mean for your sales department? Get an overview of the key dates, your business's role, transparency and how to assess specific AI tools.

Author: Agent360 editorial team
Published: 5 June 2026
Last updated: 23 October 2026
Reading time: 19 min read

What do Danish businesses need to know about the EU AI Act?

The EU AI Act is an EU regulation that governs AI according to its use and the risk it carries. For a sales department, the first step is to map the tools, what they are used for and the business's role.

The duty to disclose AI interaction has applied since 2 August 2026. The fine ceiling for Article 50 on transparency is up to EUR 15 million or 3% of global annual turnover. The ceiling of EUR 35 million or 7% applies to breaches of the prohibitions in Article 5. See the detailed conditions in the section on fines below.

TL;DR:

  • The duty to disclose AI interaction has applied since 2 August 2026, while the requirements for high-risk systems in Annex III apply from 2 December 2027
  • Assess each tool's purpose and your role; a sales product doesn't have one fixed risk class across all uses
  • The fine ceiling depends on the breach: Article 50 on transparency falls under the EUR 15 million / 3% ceiling, while EUR 35 million / 7% applies to the prohibitions in Article 5

Timeline updated: October 2026

The EU AI Act is an EU regulation (Regulation (EU) 2024/1689) that establishes a risk-based set of rules for the development, supply and use of AI systems in the EU. The regulation applies directly in all member states (including Denmark) without any need for national implementing legislation for the regulation itself.

What is the EU AI Act, and why does it affect the sales department?

The regulation entered into force on 1 August 2024, and its requirements become applicable in phases. It can be relevant when a business develops, supplies or uses AI. Which duties apply depends, among other things, on the business's role, the system's purpose and the scope of the rules.

According to the European Commission's official page on the AI Act, the aim is to ensure that AI systems on the EU market are safe and respect fundamental rights.

A sales department can use AI for lead scoring, CRM updates, conversation analysis, calls, forecasts and email drafts. Map the tools you actually have, including the ones individual employees have started using on their own.

Keep the requirements of the AI Act and the data protection rules separate in your review. EU hosting or a consent policy does not, on its own, show that a solution meets both sets of rules.

How does the AI Act classify the different risk categories?

The table is an overview. Classifying a specific system requires an assessment of its intended use; obligations can overlap.

EU AI Act risk categories and relevance for sales AI (2026)

Risk category Definition Examples in sales Requirements Fine level
Prohibited (unacceptable risk) Practices covered by Article 5, subject to the provision's specific conditions and exceptions For example, certain forms of manipulation or social scoring Prohibition, where the use is covered Up to EUR 35 million / 7% of turnover
High risk Systems that meet Article 6, including uses listed in Annex III For example, creditworthiness of natural persons or screening of job applicants, following a specific assessment Requirements depend on role and use; providers and deployers have different duties Up to EUR 15 million / 3% of turnover
Limited risk Certain AI systems with transparency obligations under Article 50 For example, direct AI interaction and certain forms of synthetic content Disclosure or labelling under the relevant provision and its exceptions Up to EUR 15 million / 3% of turnover
Minimal risk Systems outside the prohibitions in Article 5 and the high-risk rules; any transparency requirements are assessed separately For example, administrative support, following an assessment of the specific function Continue to check Article 4, Article 50 and other relevant legislation, among other things Depends on the rule that may be breached

Use Article 6 and Annex III for classification and Article 50 for the transparency requirements. A change of purpose can change the assessment. Ranking sales leads and assessing the creditworthiness of natural persons, for example, are different uses.

How do you assess typical sales AI tools?

Start with the function and the use, not the product name. What follows are questions for your review, not a pre-approval of any particular tools.

Administrative support and lead work

CRM updates, forecasts, email drafts and prioritisation of sales leads should be assessed on the basis of their specific purpose. A tool does not automatically become high-risk because it is used in sales, but the label "CRM" is not an exemption from the rules either. Also check AI literacy, any transparency requirements and the processing of personal data.

AI interaction and synthetic content

Article 50 contains different duties for, among other things, direct AI interaction, synthetic content and certain uses of biometric data. A chatbot and a recorded, transcribed conversation therefore don't necessarily need to be treated the same way.

For AI calls, you should look at how the disclosure of AI is given at the first interaction. Also read our guide to legal AI calls in Denmark 2026.

Uses that require a high-risk assessment

Annex III covers, among other things, assessing the creditworthiness of natural persons, recruitment and certain forms of employee evaluation. The provisions have limits; for example, not every voice function is remote biometric identification.

Clarify the classification under Article 6 and then the business's role. Provider duties under Article 16 and deployer duties under Article 26 are different.

Which key AI Act dates should Danish businesses know?

Here are the key application dates. Transitional rules and the specific use must be part of your assessment.

Phase 1: 2 February 2025 (ALREADY APPLIES) From this date, among other things, the prohibitions on certain forms of social scoring, manipulation and exploitation of vulnerability under Article 5 apply, subject to the provision's specific conditions and exceptions. Certain newer prohibitions in Article 5 apply only from 2 December 2026, cf. Article 113. Article 4 also requires providers and deployers to take measures that support the development of AI literacy among employees and others who use AI on their behalf. The measures must take into account, among other things, experience, education, the context of use and the people affected; the duty does not require a guarantee of a particular individual skill level.

Phase 2: 2 August 2025 (ALREADY APPLIES) The rules for providers of general-purpose AI models (GPAI) become applicable, with transitional rules for models that were already on the market. The duties that apply to a model provider must be kept separate from the sales department's duties as a user of a tool.

Phase 3: 2 August 2026 (ALREADY APPLIES) The duty to disclose AI interaction (Article 50(1)) applies from this date. The requirements for high-risk systems in Annex III were postponed to 2 December 2027 by Regulation (EU) 2026/1744 of 8 July 2026. Providers of AI systems, including general-purpose systems, that generate synthetic audio, images, video or text and that were placed on the market before 2 August 2026 must take the necessary steps to comply with Article 50(2) by 2 December 2026 at the latest.

Phase 4: 2 August 2028 The date applies to Article 6(1) and the corresponding obligations for the high-risk systems the provision covers.

Important: In November 2025 the European Commission presented a "Digital Omnibus" package. The AI part has since been adopted as Regulation (EU) 2026/1744 of 8 July 2026. It postpones the high-risk obligations in Annex III to 2 December 2027 and Annex I to 2 August 2028. The duty to disclose AI interaction under Article 50(1) has applied since 2 August 2026. Article 50(2) has a separate transitional deadline for certain existing systems, as described above.

What is the Danish Data Protection Agency doing in practice in 2026?

Supervision of the AI Act is divided by rule and sector. The Danish Data Protection Agency supervises the data protection rules and has been designated for selected AI prohibitions; it is not the general supervisory authority for all AI systems. The Danish Agency for Digital Government is the central point of contact and coordinates the work.

The relevant authority depends on the specific rule:

  • Data protection: The Danish Data Protection Agency supervises GDPR
  • Selected AI prohibitions: The Danish Data Protection Agency's designation covers Article 5(1)(d) and (g)
  • Coordination: The Danish Agency for Digital Government acts as the central point of contact for the AI Act and as the market surveillance authority for Article 5(1)(a)-(c), (e) and (f)
  • Other rules: Identify the relevant authority from the provision and the sector in the current supervisory model

The Danish Data Protection Agency's published focus for 2026 includes, among other things, AI-based decision support in the healthcare sector and AI used for monitoring and control of citizens in care. The regulatory sandbox, which the Danish Data Protection Agency runs together with the Danish Agency for Digital Government, offers guidance on GDPR and risk classification under the AI Act. Taking part is not an approval of a solution.

See the Danish Data Protection Agency's guidance and the Danish Agency for Digital Government's overview of the supervision for the specific division of authority.

What fine ceilings does the AI Act set?

Article 99 of the AI Act sets different fine ceilings depending on the nature of the breach. Here are three key categories:

Tier 1: Prohibited AI practices Up to EUR 35 million or 7% of global annual turnover in the preceding financial year for breaches of the prohibitions in Article 5. As a starting point, the higher amount applies; special rules for business size are described below.

Tier 2: High-risk requirements and transparency, among other things Up to EUR 15 million or 3% of global annual turnover in the preceding financial year for the obligations referred to in Article 99(4). This includes, among other things, certain duties for high-risk systems and the transparency requirements in Article 50.

Tier 3: Incorrect information Up to EUR 7.5 million or 1% of global annual turnover in the preceding financial year for incorrect, incomplete or misleading information given to notified bodies or national competent authorities in response to a request.

The amount ceiling and the percentage ceiling are, as a starting point, taken together, so the higher one applies. For SMEs the lower ceiling applies in all three categories. For small mid-caps (SMCs) the lower ceiling applies in categories 2 and 3. Turnover alone determines neither the business category nor the specific fine, which depends on the circumstances of the case. See Article 99.

The fine ceilings are upper limits. They do not in themselves give a date for when a particular breach will lead to an enforcement case.

How does your sales department prepare in practice?

Here is a concrete action plan in four steps for Danish sales organisations.

Step 1: Map your AI systems

Note the vendor, purpose, input, output and responsible employee. Include lead work, conversation analysis, administration and coaching. Describe in particular whether the output is used for decisions about customers or employees.

Step 2: Review the transparency requirements

Clarify the relevant parts of Article 50 for each system. The duty under paragraph 1 rests on the provider; as a deployer you should check that the solution gives the relevant disclosure, and settle the allocation of roles in the agreement. For AI calls, a clear opening can be: "You are speaking with an AI assistant from [Company]." That is a practical example, not documentation that all the requirements for the system are met.

Also read our article on legal recording of sales conversations.

Step 3: Clarify responsibility for high-risk systems

A provider is responsible for, among other things, the relevant conformity assessment before placing on the market or putting into service. A business that uses a high-risk system as a deployer has other duties, including use in accordance with the instructions, human oversight and monitoring of operation. See Article 16 and Article 26.

So clarify the role and the documentation needs before you allocate tasks. Don't use a standard timetable as proof that your review is sufficient.

Step 4: Build AI literacy (ongoing)

Article 4 requires measures that support the development of AI literacy, without requiring a guarantee of a particular individual level. For the sales department, relevant measures could include, for example:

  • Training in what the AI tools do and don't do
  • Understanding when AI output needs to be verified
  • Knowledge of the business's AI policy and escalation procedures

Adapt the measures to employees' backgrounds, the specific AI systems and the people the use affects.

How does the AI Act differ from GDPR, and how do they overlap?

GDPR regulates the processing of personal data. The AI Act regulates AI systems and models with requirements that depend, among other things, on role and use. Article 2(7) states that the data protection rules continue to apply.

A GDPR review therefore doesn't replace the assessment under the AI Act. Conversely, an AI classification doesn't document a legal basis for processing, deletion practice or lawful data transfers. Bring the documentation together, but assess the duties separately.

For a detailed review of data sovereignty and AI in sales, see our article on data sovereignty.

What does the AI Act mean for voice agents and AI calls?

AI calls can make Article 50 relevant. Under paragraph 1, the provider must ensure that people are informed of direct AI interaction, unless it is obvious to a reasonably well-informed, observant and circumspect person in the specific context. The provision also contains a separate exception for certain purposes of detecting and preventing crime that are authorised by law.

A clear disclosure at the start of the conversation is a practical starting point. It does not replace the assessment of the other requirements for calls and data processing.

Practical checkpoints for the agreed solution:

  • Open the conversation with: "Hello, you're speaking with [Name], an AI assistant from [Company]"
  • Offer the option of being transferred to a person
  • Document that the disclosure duty has been met

Emotion recognition: Emotion recognition under the AI Act concerns identifying or inferring emotions or intentions from biometric data. Article 5(1)(f) prohibits inferring emotions in the workplace and in educational institutions, with exceptions for medical or safety purposes. Service improvement is not in itself an exception. If a use of an emotion recognition system is legal, the deployer has its own duty under Article 50(3) to inform the people affected about how the system operates. The label sentiment analysis therefore does not determine legality; the specific function, data and use must be assessed. See Article 5 and the definition in Article 3(39), as well as the further interpretation in recital 18.

Ask the vendor for documentation of transparency, the legal basis for processing and data processing in the specific solution. Read more about AI calls and Danish legislation in 2026.

Which assumptions should Danish businesses check?

  • "GDPR already covers us": Review the AI Act separately.
  • "We know all our tools": Ask your employees and go through the actual work processes.
  • "The vendor has classified the system": Check that the assessment covers your use.
  • "A person can always step in": Test who can stop or change a decision, where relevant.
  • "The documentation exists": Find the actual instructions, agreements and assessments, and agree who keeps them up to date.

Can documentation help in the sales process?

Documentation can make it easier to answer a customer's specific questions about AI use and data. Show what you have looked into, which limitations apply and who is responsible. Avoid using "AI Act-compliant" as a promise without qualification.

A practical start is to gather vendor information, purpose, risk assessment and responsibilities in one place. That gives you a basis for the review, but it is not in itself proof of legality.

What should be on your AI Act compliance checklist?

Here is the practical checklist for Danish sales organisations:

  • AI inventory created: All AI systems in the sales department mapped with name, vendor and purpose
  • Classification assessed: Purpose, role and relevant provisions are documented
  • Transparency reviewed: The relevant requirements and exceptions in Article 50 have been assessed
  • Responsibility clarified: Provider and deployer tasks are allocated for any high-risk systems
  • AI literacy measures in place: The measures are adapted to employees' backgrounds, the use and the people affected
  • Technical documentation: Clarified whether you are the provider of a high-risk system and, if so, who maintains the documentation
  • Vendor agreements reviewed: The documentation covers the specific use and the allocation of responsibility
  • Sandbox assessed: Clarified whether you or your vendor can apply to the regulatory sandbox. The AI Act's scheme is aimed at providers or prospective providers; as a deployer you can apply together with the relevant provider
  • Incident response plan: Procedure for handling AI errors or complaints
  • Ongoing monitoring: Process for regularly reviewing the compliance status of AI systems

Key sources and resources

Frequently asked questions about the EU AI Act and sales

Q: Is AI-based lead scoring high-risk under the AI Act?

Not automatically. Assess the purpose and the specific use under Article 6 and Annex III. Prioritising sales leads and assessing the creditworthiness of natural persons are different uses. There may also be duties other than the high-risk requirements, including data protection and AI literacy.

Q: Does my AI voice agent have to say that it is an AI?

Article 50(1) requires, as a starting point, disclosure of direct AI interaction and contains an exception where it is obvious in the specific context. For a sales call, a clear disclosure in the opening is a practical starting point. Also check the other requirements for the call; the disclosure alone does not document that the whole solution is legal.

Q: Does the AI Act only apply to businesses in the EU?

No. Article 2 covers, among others, providers that place AI systems or GPAI models on the EU market, even if they are established outside the EU. It also covers certain providers and deployers in third countries where the system's output is used in the EU. The scope has exceptions, which must be assessed case by case.

Q: What happens if we use ChatGPT or Claude in the sales department?

Distinguish between the model provider and the business's own role. Using an AI system under the business's responsibility can make the business a deployer. Developing or commissioning a system that is marketed or put into service under its own name can also make the business a provider. Clarify the role under Article 3 and the specific use; it is not determined by which model sits underneath alone.

Q: What is the Regulatory Sandbox for AI in Denmark?

The regulatory sandbox is a collaboration between the Danish Data Protection Agency and the Danish Agency for Digital Government on guidance about GDPR and risk classification under the AI Act. The AI Act's scheme is aimed at providers or prospective providers; a deployer can apply together with the relevant provider. It grants neither exemption from the rules nor advance approval of an AI solution. The authorities do not provide technical infrastructure or a test platform; participants are themselves responsible for documentation and technical support.

Q: Can we use AI for conversation analysis and recording under the AI Act?

It depends on the function, the data basis and the use. Inferring emotions from biometric data in the workplace and in educational institutions is covered by Article 5(1)(f), with exceptions for medical or safety purposes. Service improvement is not a general exception. Transparency and a valid legal basis for processing do not in themselves make a prohibited use legal; other forms of conversation analysis must also be assessed under the relevant rules. Read our detailed guide to legal recording of sales conversations with AI and GDPR for the full review.

Q: How should we approach enforcement of the AI Act?

The application dates of the rules do not fix the date of a particular fine or enforcement case. Assess the duties that apply to your specific systems and role, and follow the guidance and decisions of the relevant authorities. The Danish Data Protection Agency has a limited AI supervisory remit, while the Danish Agency for Digital Government coordinates the Danish effort.

Conclusion: Start with the specific AI systems

Map the tools, clarify the purpose and role, and allocate the tasks. Keep the documentation for the AI Act and data protection together, so you can follow up on changes in systems and use.

Contact Agent360 if you'd like to discuss a specific AI workflow in the sales department.

Tags: #EU AI Act #compliance #GDPR #AI regulation #sales #Danish legislation #Danish Data Protection Agency #AI risk categories

Unsure about the rules for your own set-up?

We look at what you use today and what it requires: documentation, consent and what the EU AI Act means for you. 20 minutes, no commitment.

How we built it · Sales & follow-up

Sales cockpit with AI

A call list for salespeople. The call is transcribed in Danish, AI suggests the outcome (e.g. “Call again”) and drafts the email based on what the customer said.

  1. The salesperson calls
  2. AI writes transcript and email
  3. The salesperson sends
See the call and the follow-up →