Part of the AI Regulation News hub.
Azerbaijan added two deepfake offences to its Criminal Code and a general labelling duty for AI-made material shown to the public
Two of the four amendments are criminal. The one most lawyers will actually have to advise on is the fourth, a marking duty that is not limited to material made without consent.
Bottom line: Binding. This is an adopted law signed by the President, not a draft or a consultation. The e-qanun record carries the status Quvvededir, meaning in force.
Who this affects: Defence counsel and media lawyers in Azerbaijan, in-house counsel at media outlets registered under the Media Law, and platform counsel handling Azerbaijani takedown demands.
Issue date: 21 April 2026, signed in Baku by President Ilham Aliyev as Law No. 387-VIIQD. The amending text sets no commencement clause of its own; the e-qanun register records entry into force on 6 May 2026 and no compliance deadline.
What changed: New Criminal Code Articles 148-2 and 242-1; Article 148-2.1 added to Criminal Procedure Code Article 37.3; new Articles 13-2.3.9-1 and 13-2.7 in the Information Law; new Article 14.1.13-1 in the Media Law with two cross-reference updates.
Analysis: The criminal articles both turn on the absence of consent. The labelling duty in new Article 13-2.7 does not. It attaches to any AI-made photo, video or audio material that is publicly displayed, consented to or not, which makes it the provision with the widest reach and the one with no stated penalty in this law.
Primary sources: Law 387-VIIQD, full text (e-qanun.az, Azerbaijani) · Criminal Code of the Republic of Azerbaijan (e-qanun.az) · Law on Media, 2021 (e-qanun.az)
- Instrument (EN)
- Law amending the Criminal Code, the Criminal Procedure Code, the Law on Information, Informatization and Protection of Information, and the Law on Media
- Number
- 387-VIIQD
- Authority
- Milli Majlis of the Republic of Azerbaijan; signed by President Ilham Aliyev
- Jurisdiction
- Azerbaijan
- Status
- Adopted and published on the official e-qanun register, marked in force
- Bindingness
- Binding statute. Criminal penalties attach to Articles 148-2 and 242-1
- Issue date / next deadline
- Adopted 21 April 2026; in force 6 May 2026 per the e-qanun register. No transition period or compliance deadline appears in the text
- Instruments amended
- Four: Criminal Code, Criminal Procedure Code, Information Law, Media Law
- Primary source
- https://e-qanun.az/framework/61783
The two new criminal articles
Article 1 of the law inserts Article 148-2 into the Criminal Code. It covers making photo, video or audio material that does not reflect reality by means of artificial intelligence technologies or special software, using a person's image or voice without that person's consent, and it covers distributing that material in the media or, where it is publicly displayed, on internet information resources or in information and telecommunication networks.
The base penalty is a fine of three thousand to seven thousand manat, or community work of 360 to 480 hours, or restriction of liberty for up to three years, or deprivation of liberty for the same term.
Article 148-2.2 raises the penalty to three to five years of deprivation of liberty in four cases: commission by a group or by a group acting in prior concert, commission against two or more persons, commission for the purpose of tarnishing a person's honour and dignity or discrediting them, and commission against a victim or the victim's close relatives in connection with the performance of official duties or public duty.
Article 242-1 is the sexual-content counterpart. Same technological means, same consent condition, pornographic or sexual material rather than material that merely does not reflect reality. The penalty there is three to seven years of deprivation of liberty, with no fine and no community-work alternative.
A labelling duty that does not depend on consent
Article 3 of the law amends the Information Law. It adds Article 13-2.3.9-1, which lists the same nonconsensual AI material as a category of content, and it adds Article 13-2.7, which is a different kind of provision altogether.
New Article 13-2.7 says that where photo, video or audio material prepared by means of artificial intelligence technologies is publicly displayed, it must be provided with marking that clearly shows, and that is easily visible, that it was prepared using such technologies.
Nothing in that sentence limits the duty to material made without consent, to synthetic depictions of real people, or to any particular platform. It is a public-display trigger. The law as I read it does not attach a penalty to Article 13-2.7 itself, and I have not opened the Code of Administrative Offences to see whether one sits there.
Procedure and the Media Law
Article 2 makes a single procedural edit: in Criminal Procedure Code Article 37.3, the figures 148-2.1 are inserted after 148-1. The law does not restate what Article 37.3 governs, and I did not open that article, so I am not characterising the effect.
Article 4 amends the Media Law. New Article 14.1.13-1 prohibits the dissemination of the nonconsensual AI material described above. Article 41.1.1 and Article 66.1 are then updated to cite the new provision, which is how the Media Law's existing consequences reach it.
Read together, the four articles do the same work three times over: a crime for the worst conduct, a content prohibition for media entities, and an information-law listing for online resources.
Where the drafting leaves room
The phrase carrying the most weight in Article 148-2 is material that does not reflect reality. Neither that phrase nor artificial intelligence technologies is defined in the amending law, and the amending law is a set of insertions, so any definitions have to come from the host statutes.
The consent element is doing a lot of work as a limiting principle. Satire, dubbing, restoration and reconstruction all involve a person's image or voice, and the criminal articles turn on whether the person agreed, not on what the maker intended.
The marking obligation and the criminal offences also have different addressees. The criminal articles reach whoever makes or distributes. Article 13-2.7 is written in the passive and attaches to the material at the point of public display.
What we did not verify
I opened the full text of Law 387-VIIQD on the official e-qanun.az register, including all four amending articles, the signature block naming President Ilham Aliyev, the place Baku, the date 21 April 2026 and the number 387-VIIQD.
I did not open the consolidated Criminal Code, the Criminal Procedure Code, the Information Law or the Media Law, so I have not read Article 37.3, Article 41.1.1 or Article 66.1 in their own text. I did not open the Code of Administrative Offences and I did not find an official English translation of this law.
I am not claiming what Criminal Procedure Code Article 37.3 does, whether Article 13-2.7 carries any sanction, whether the law applies to content published before 21 April 2026, or how any of this has been enforced.
If you advise anyone publishing AI-made audio or video into Azerbaijan, the criminal articles are the headline and the labelling rule is the operational problem. Article 13-2.7 applies to AI-made material that is publicly displayed without asking whether anyone was depicted or whether anyone consented. Build the marking into the asset rather than into the caption, since the duty says the marking must be easily visible on the material shown.
Source File
https://e-qanun.az/framework/61783
Open e-qanun.az/framework/61783 and confirm three things: the signature block reads Baku, 21 aprel 2026-ci il, No 387-VIIQD; Article 1 inserts both 148-2 and 242-1 with the fine band of three thousand to seven thousand manat in 148-2.1; and Article 3 inserts 13-2.7 with the public-display marking duty.
Where photo, video or audio materials prepared by means of artificial intelligence technologies are publicly displayed, they must be provided with marking that clearly shows, and that is easily visible, that they were prepared using such technologies. ยท Law 387-VIIQD, new Article 13-2.7 of the Information Law, TLY working translation from the Azerbaijani, 21 April 2026
FAQ
Does the labelling duty only cover deepfakes of real people?
Not as drafted. New Article 13-2.7 of the Information Law applies to photo, video or audio material prepared by means of AI technologies when it is publicly displayed. The consent condition and the reference to a person's image or voice appear in the criminal articles, not in 13-2.7.
What is the maximum sentence?
Seven years of deprivation of liberty, under new Criminal Code Article 242-1, for making or distributing pornographic or sexual AI material using a person's image or voice without consent. Article 148-2.2 tops out at five years.
Is a fine available for the sexual-content offence?
The text of Article 242-1 sets only deprivation of liberty of three to seven years. The fine, community work and restriction-of-liberty alternatives appear in Article 148-2.1, not in 242-1.
When does the law take effect?
The amending text carries the date 21 April 2026 and contains no commencement clause of its own. The e-qanun register supplies the rest: it marks the law Quvvededir, in force, and records an entry-into-force date of 6 May 2026. There is no transition period and no compliance deadline in either the text or the register.
Related briefings
Sponsored Training
Practical AI training for regulated professionals, built around verification, documentation and a defensible process. See the courses.