New requirements for labelling AI-generated marketing: What does Article 50 of the EU AI Act mean for asset managers?

The EU AI Act entered into force on 1 August 2024. It is one of the first comprehensive sets of rules regulating the development, provision, and use of AI. It is intended to ensure that AI systems are used in a safe, transparent, and responsible manner.

Different rules apply in stages. From 2 August 2026, the requirements for labelling AI-generated content applied. Article 50 imposes several separate obligations that apply to different actors and different types of content. The rules distinguish between two actors: providers and deployers.

Providers of AI systems, such as Microsoft, that generate synthetic content (including general-purpose AI systems) must ensure that AI-generated or manipulated audio, image, video, and text output is marked in a machine-readable format and detectable as artificially generated or manipulated.

Deployers, such as asset managers, have a narrower, separate disclosure duty under Article 50(4). They must disclose that content is artificially generated or manipulated only where it constitutes: (i) a deep fake (i.e., AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful); or (ii) AI-generated or manipulated text published with the purpose of informing the public on matters of public interest. An exception applies to (ii) where the content has undergone human review or editorial control, and a natural or legal person holds editorial responsibility for its publication. Superficial checks such as spell-checking are not sufficient.

For asset managers, the disclosure duty under Article 50(4) will in practice turn almost entirely on alternative (ii), i.e., AI-generated or manipulated text informing the public on matters of public interest, since ordinary fund marketing materials, fact sheets and investor communications do not typically involve deep fakes within the meaning of alternative (i). The disclosure duty will generally not apply to ordinary fund marketing materials, fact sheets and investor communications produced with AI assistance (such as Copilot), provided the content does not constitute AI-generated or manipulated text published to inform the public on matters of public interest without genuine human review or editorial control.

The guidelines published by the European Commission clarify that the notion of "matters of public interest" under Article 50(4) cover matters relevant to society at large, at a local, national, EU or international level, that merit public debate or scrutiny, including politics and democratic processes, public administration and services, the administration of justice and law enforcement, fundamental rights, public security, public health, environmental protection, consumer safety, and any economic, financial, political, scientific or cultural development that may be a relevant subject of public debate.

This is directly relevant for asset managers. The guidelines give AI-manipulated corporate reports containing investor information published on a listed company's website as an example of text that does inform the public on a matter of public interest. By contrast, AI-manipulated advertising or product description text is given as an example that falls outside scope, provided it does not include claims relating to matters such as health, consumer safety, or sustainability. This suggests that ordinary fund marketing collateral, fact sheets and generic investor communications will typically sit outside Article 50(4), but investor-facing disclosures with a public-interest character (for example, sustainability-related disclosures) may fall within it unless the human review/editorial control exception applies.

Asset managers should identify whether any AI-generated or AI-assisted content they produce or publish qualifies as AI-generated or manipulated text published to inform the public on matters of public interest without genuine human review or editorial control; in practice, this is the alternative of Article 50(4) relevant to this sector, since fund marketing and investor communications do not typically involve deep fakes. Only such content triggers the disclosure duty under Article 50(4); the mere fact that content was produced with AI assistance does not, by itself, trigger a disclosure obligation for asset managers as deployers.

Furthermore, asset managers should integrate a disclosure process into their existing workflows for fund marketing and client communication production but scoped to the content caught by Article 50(4) (AI-generated/manipulated text informing the public on matters of public interest without genuine human review or editorial control). Asset managers should establish internal guidelines for identifying such content and for when and how it must be disclosed as AI-generated or manipulated, rather than applying disclosure requirements to AI-assisted content generally. This applies irrespective of whether material is produced in-house or through external partners (including advertising agencies, content creators, and robo-advice platform providers), and asset managers should ensure that such partners are familiar with the requirements and that responsibility for disclosure is clearly allocated, including contractually where relevant.

Breach of the transparency requirements in Article 50 can result in significant sanctions. The Regulation establishes a sanctions regime under which breaches of Article 50 can result in administrative fines of up to EUR 15 million or up to 3% of the company's total global annual turnover, whichever is higher. The national supervisory authorities are responsible for enforcement.

In addition to direct fines, failure to disclose AI-generated content in fund marketing or investor communications can cause reputational damage for assets managers and undermine confidence among investors and distribution partners.

1. Map the use of AI in fund marketing and investor communications. 
   Asset managers should review and map the extent to which AI tools are used to generate or edit fund marketing material, fact sheets, investor presentations, performance commentary,  and robo-advice outputs, focusing on identifying any content that may constitute AI-generated/manipulated text informing the public on matters of public interest. 
    
2. Adopt internal guidelines and processes for disclosure. 
   Internal policies must allocate responsibility for when and how AI-generated content must be disclosed. Integrate the disclosure requirements into approval workflows.
    
3. Impose requirements on distributors and technology partners. 
   Ensure that providers, distributors, advertising agencies, and content creators are familiar with and comply with the requirements for disclosing AI-generated content. Consider including specific compliance obligations in distribution, platform, and cooperation agreements.
    

The Digital Omnibus on AI provides a grace period until 2 December 2026 for the machine-readable marking and technical detection requirements under Article 50(2), but this concession applies only to providers of generative AI systems already placed on the market before 2 August 2026. It does not affect the deployer disclosure duty under Article 50(4), which applies to asset managers from 2 August 2026 without any grace period.