Romania - Responsible Technology Use (PL-x 471/2023)

Overview

PL-x No. 471/2023 (Senate file L295/2023; Chamber PL-x 471/2023) is a legislative initiative introduced in 2023 to regulate the use and dissemination of audio, image and video content created or altered by artificial intelligence and related technologies in ways that could mislead the public. The project, described in the Senate legislative dossier and an explanatory memorandum, frames deepfakes as a rapidly developing technological risk that can cause reputational harm, fraud, privacy violations and threats to public order. The bill requires explicit, visible and audible notices on deepfake content, assigns monitoring and enforcement powers to the Consiliul Național al Audiovizualului (CNA), and establishes administrative and — in more contested amendments considered at Chamber level — penal sanctions. Official project documents (Senate legislative record and PDFs of the form adopted by the Senate and the explanatory memorandum) are available in the Senate legislative archive. See the legislative file and supporting documents: Senate legislative file L295/2023 and the forma adoptată PDF: Text (Senate - forma adoptată).

Definitions

The bill defines key terms for the regime. "Deepfake" is defined as any image, audio and/or video content fabricated or manipulated—typically via AI, VR or AR—so as to create the appearance that a person said or did things they did not in reality say or do, particularly where consent is absent. The proposal also distinguishes between legitimate uses (artistic, advertising, film-making, research) and malicious uses (fraud, extortion, disinformation, impersonation). The text sets the scope for "dissemination" and "distribution" (online platforms, broadcast media, social networks) and designates obligations for authors, distributors and hosting platforms.

Governance and Institutional Framework

The bill vests primary supervisory and enforcement authority with the Consiliul Național al Audiovizualului (CNA) for audiovisual distribution and broadcast, and foresees coordination with other state institutions, including data-protection authorities and law-enforcement bodies. The Senate dossier includes formal aviz documents from the Consiliul Legislativ and the CNA; copies of those opinions and the explanatory memorandum are published on the Senate site (for example, the CNA opinion PDF and the Consiliul Legislativ opinion are linked on the legislative dossier). The CNA is empowered to investigate complaints (initiated by victims or ex officio), order removal or suspension of improperly labelled material, and impose administrative sanctions. The text contemplates collaboration arrangements with national R&D institutions for technical detection assistance. The legislative file and formal opinions are available here: CNA opinion (Senate PDF), Consiliul Legislativ opinion (Senate PDF) and Explanatory memorandum (Senate PDF).

Key Focus Areas

The bill concentrates on several complementary policy streams: transparency and labelling, consumer and victim protection, market supervision and platform obligations, criminal/administrative deterrence and coordination with EU law. Transparency: the law mandates that deepfake audiovisual content carry a visible visual label covering a specified minimum share of the image area and/or an audible notice at the start and end of audio/video streams stating that the material is artificially generated or manipulated. Victim protection: the regime enables individuals to file complaints and seek rapid removal orders where content portrays them without consent; the CNA may order removal and oblige authors to cease dissemination. Platform and broadcaster duties: broadcasters and platform operators are required to apply labelling standards, keep records to enable identification of creators (subject to data-protection constraints), and cooperate with enforcement requests. Market surveillance and conformity: the bill contemplates registration/notification for professional uses (advertising, film, marketing) so that legitimate commercial uses are identifiable. Deterrence: the proposal sets administrative fines for labeling failures and — in some Chamber-stage amendments that prompted public criticism — criminalizes the malicious creation or repeated dissemination of deepfake content, with specified custodial penalties. The legislative dossier emphasizes proportionality but the introduction of penal sanctions has been controversial among civil-society groups and legal experts.

Implementation Framework

Implementation responsibilities are distributed across the CNA (operational monitoring and sanctions for audiovisual media), law enforcement (investigation of fraud, extortion, criminal impersonation under the Criminal Code), and data-protection authorities for processing and retention concerns. The bill foresees procedural rules for complaint filing, notice-and-takedown by regulated broadcasters, and potential referral to prosecutors when criminal offences are detected. The CNA may adopt, by decision, technical formatting rules for visible labels and audible notices, and may enter technical cooperation agreements with national research institutions for detection tools. The project envisages transitional provisions for legacy content and includes carve-outs for certain clearly marked and registered uses (commercial/film productions), although civil-society stakeholders have asked for broader express exemptions for satire, parody and research to protect freedom of expression.

Monitoring and Evaluation

Monitoring will be led by the CNA with authority to act on complaints and to initiate ex officio procedures; the CNA can require removal from broadcast and issue fines for violations. The bill requires reporting from regulated entities and contemplates periodic public reporting by the CNA on enforcement actions and trends. The legislative dossier contemplates technical monitoring cooperation with research entities for improved detection; the law would also require regulatory review after an initial period to assess effects on rights and information integrity and to align with EU instruments such as the AI Act and DSA.

Penalties, Liability, and Appeals

The proposal establishes an administrative sanctions regime (fines) for failure to label deepfake material and for non-compliance with CNA orders; the bill in amended Chamber form reportedly also introduces criminal penalties for malicious creation and dissemination of deepfakes in particular circumstances. Administrative sanctions include graduated fines (first offence and repeat offences with increased amounts) and the CNA's power to stop broadcasts and require content removal. Where criminal offences are alleged (fraud, extortion, aggravated deception), the bill contemplates referral to prosecutors; it also raises questions regarding proportionality and safeguards for lawful expression, prompting civil-society requests for exceptions and clearer definitions to avoid chilling effects.

Relationship to Other Instruments

PL-x 471/2023 intersects with Romanian criminal law provisions (fraud, falsification, data integrity), national audiovisual legislation, personal-data protection rules, and pending EU-level frameworks. The bill's authors mention international precedents and aim for complementarity with EU regulation; however, critics emphasize the need to avoid duplication or conflict with the EU Artificial Intelligence Act and the Digital Services Act. The Senate legislative file includes formal opinions from the Consiliul Legislativ and CNA which discuss compatibility and constitutional considerations.

International Alignment

The explanatory memorandum and public debates explicitly reference EU-level regulatory activity (the AI Act and the Digital Services Act) and jurisprudence on freedom of expression and proportionality. Stakeholders recommend harmonization with the AI Act's definitions and risk-based approach and alignment with DSA obligations for platforms. International human-rights and free-expression standards have been cited in critiques seeking narrow, targeted restrictions on malicious conduct while preserving legitimate uses of synthetic media.

Implementation Timeline

Sources and References