Guidelines on the implementation of the transparency obligations for certain AI systems under Article 50 of the AI Act — C(2026) 5054 final, Annex

(46) Providers of interactive AI systems are exempted from the transparency obligation under Article 50(1) AI Act if they are authorised by law to detect, prevent, investigate, or prosecute criminal offences, subject to appropriate safeguards for the rights and freedoms of third parties.

(47) ‘Authorised by law’ is to be understood to comprise Union law, as well as Member States’ national law adopted in compliance with Union law. The law authorising the use of the AI system must clearly specify the purposes and circumstances of permitted use and provide for appropriate safeguards to protect the rights and freedoms of third parties. Considering the technology neutral approach of the AI Act, it is not necessary to identify or authorise specific AI systems by law where the use of the AI system is generally envisaged within the scope of the legally authorised activity and complies with the applicable legal framework.

(48) To fall within this exception, the purpose of the AI system must be to detect, prevent, investigate, or prosecute criminal offences (e.g. AI-undercover agent, AI systems used under human control of investigators for online investigations or communication with persons who may be involved in criminal activities). The exception is not restricted to the use of such AI systems only by law enforcement authorities as defined in Article 3(48) AI Act, but may also cover interactive (or generative) AI systems used by other Union or national public authorities, so long as their use is authorised by law to detect, prevent, investigate, or prosecute criminal offences, and subject to appropriate safeguards to protect the rights and freedoms of third parties. The exemption does not require the AI system to be capable of conducting the investigations independently. If a respective system can both be used for law enforcement purposes, as well as for other regular purposes, the applicability of the exemption should be assessed in light of the particular deployment. Transparency obligations continue to apply where the same AI system is deployed for other purposes than the legally authorised detection, prevention, investigation or prosecution of criminal offences. Transparency still needs to be provided

to natural persons when they are interacting with the system in the context of its regular purposes.

(49) The exception does not apply if (i) the AI system is available to the public, and (ii) the system offers a functionality to individuals to report criminal offences. Such systems may be used, but they remain subject to the transparency obligation under Article 50(1) AI Act, so natural persons must be informed they are interacting with such systems. For greater clarity, this exception concerns solely the obligation under Article 50(1) to inform natural persons that they are interacting with an AI system (22). It is without prejudice to applicable Union and national law, including procedural rules concerning the documentation, disclosure, admissibility and assessment of evidence. Examples of AI systems made available to the public to report criminal offences that are not exempted and fall within the scope of Article 50(1) AI Act:

- AI-enabled police chatbots deployed on official police websites or mobile applications that allow individuals to report criminal offences or complaints. These systems guide users through structured questions, collect relevant information, and triage reports before forwarding them to human officers for assessment. · AI-powered telephone hotlines operated by the police or other law enforcement authority for general public inquiries.

- AI-assisted fraud reporting hotlines or digital portals operated by financial institutions or public authorities, where users can report suspected financial crimes. The AI system interacts with the user to gather details, categorise the report, and prioritise cases, while human investigators validate the information before any action is taken.

- Virtual assistants used for witness statement collection, where victims or witnesses can submit information through an AI-driven interface. Such AI systems may support multilingual input, accessibility features, and structured evidence submission, but the final assessment and investigative steps remain with human authorities.