Introduction
Until now, pro-Islamic State (IS) Artificial Intelligence (AI) propaganda materials were generated and shared online by supporters of the caliphate. Yet, none of them had ever been located within private chats and channels belonging to the IS communication infrastructure (IS online safe-havens).
Following the Crocus City Hall attack in Moscow on 26 March 2024, a user accessed the IS server on Rocketchat and posted a video generated with AI on the server’s main discussion room. From that day to 1 May, he posted five different propaganda videos employing the same production techniques, albeit with some graphic and content differences. As of now, this is the most prominent example of an IS supporter using AI. The fact that the video was posted on Rocketchat – one of the main digital safe havens of IS since 2019 – represents a clear case of a security efficiency trade-off and a deliberate choice of the user to position itself as quickly as possible within the pro-IS digital ecosystem. The attack at the Crocus City Hall in Moscow could be considered the main source of inspiration that led to the production of these videos since it is the subject of the first one. The attack was even presented as a “new beginning” in the audio message released by the IS spokesman shared by its media arm, al-Furqan. This moment may have ignited a new era of pro-IS media operators.
The insight thus delves into how pro-IS supporters use AI to spread institutional propaganda materials of the Islamic State. By analysing the content, the videos are structured around official IS propaganda, which is presented as a newsflash by an AI-generated Arabic speaker. Specifically, the result was the creation of an online version of IS mass media news broadcasts. Moreover, the usage of official IS material, such as Amaq’s bulletins, is highly relevant because the user presents himself with all the formal features of an official media house.
Furthermore, the use of material gleaned from the official IS communication stream and the choice to share the videos within an environment considered safe underlines the willingness to flaunt this content. In other words, it is a clear attempt to promote a new brand by respecting some of the socio-digital practices found within the pro-IS ecosystem. For instance, taking inspiration from the institutional material and creating a well-recognisable brand as a media house.
The only missing element is the digital bay’ah, a declaration of the alliance towards the Caliph, usually presented as a text banner. Nevertheless, the absence of this element does not exclude future publications. In the past, other media houses have declared an alliance after consolidating their position online.
From 26 March to 4 April, the user produced and posted the videos in Arabic (two of 90 seconds and one of 60 seconds). Interestingly, three different AIs methods were employed to produce the videos:
- Character (speaker) generation;
- Voice file generation. In this case, a text-to-speech AI was used, which, starting from a written text, simply reads it with pre-selected intonation and language (Arabic);
- Lips movement and connection to the audio file.
The first video represents a jihadist press review of Amaq and IS bulletins related to the Moscow attack, while the second and the third present a press review of IS’s Middle Eastern and African operations. As shown by the screenshots (Fig. 1), the first generated character wears a military uniform, while in the other two, the speaker is dressed in traditional Islamic civilian clothes (Fig. 2).
Figure 1: military-generated character.
Figure 2: civilian-generated character.
The content of the videos remained the only constant element: official claims and statements are used as overlay images (material from Amaq, IS and Al-Naba) and the synthetic voice reads the written text. In addition, it is noteworthy that, apart from the first video, the user signs the other four videos with a logo that recalls his name on Rockechat, implying that it is its own media house.
The videos and the usage of IS content
From the analysis of the communication flow, according to the pro-IS community reactions, it is a successful attempt to provide a ‘digital press review’ service of IS operations in the Middle East and sub-Saharan Africa. This thematically overlaps with the Nashir News, although they only share official IS material. The user has also collected part of these materials, in particular war news, and pasted them into the video. Moreover, the layout of the video refers to a specific communicative imperative, recalling the style of the major television news networks, such as CNN and the BBC. At the bottom of the screen, a short text overlay can be seen, accompanied on the right by the IS official logo.
Figure 3: the logo of the user in the top right-hand corner., IS logo on the bottom right.
Even though the five videos present themselves as a new form of cutting-edge media production, it is worth noting that, except for the first video dated 26 March, some of the material shows considerable discrepancies in terms of chronology and geography between the IS media centre and Amaq bulletins and the images that follow them. The institutional material presented in the videos refers to recent IS operations, while some of the images date back to 2022, 2023, and even 2017. Specifically, in the second video, images are from 2023 and 2017 (one of which cannot be identified because its characters are too generic), while in the third video, images are from 2022 and 2023. For instance, in the case of Fig.4, the user poses an image of the Islamic State Province in West Africa (ISWAP), specifically Mozambique, which follows an IS media centre bulletin referring to an operation in Nigeria. The more remarkable part is that the user was seemingly aware of this discrepancy. He cut off the lower part of some of these images to delete the overlay text, in which IS media centre and Amaq give a short caption, location, and the date of the operation.
Figure 4: image of the operation of ISWAP in Mozambique.
The reaction of the pro-IS community
The five videos shared by the user on Rocketchat have garnered positive and favourable opinions, as shown in Fig. 5. Furthermore, an indication that the video caught the attention of the community is the presence of several pieces of advice given by other users in order to make the video theologically coherent.
Figure 5: a user asks in Arabic which agency produced the video posted on the 26 of March. The poster replies with his nickname, suggesting that whoever is behind the user identifies themselves as a spontaneous media house.
One of the community reactions was highly interesting, particularly given its possible future implications. While different users suggested some generic corrections, the most interesting interactions pertained to the videos’ language and religious elements. After receiving a request from a user for an English version, the creator of the AI video expressed positively, suggesting that they might produce new videos in English in the future. Another user was more concerned about the theological implications of using AI. Its interactions with the AI creator pushed him to blur the face of the speaker (Fig. 6) in order to not be considered haram (sinful).
Figure 6: image of the correction made by the user to comply with the community rules
This is closely tied to the potential of this new form of AI-generated content, which promises a significant increase in the usability of content produced by both IS and official media outlets. By creating a press review of the main IS news and content (the last video is based around articles from al-Naba, an Islamic State official magazine), it is assumed that the user, through its media house, is seeking to propose a new broadcast-style propaganda tool, speeding up and facilitating the dissemination of information and propaganda material to wider Western audiences.
As to how this case might represent the use of AI is still shrouded in a halo of perplexity concerning its permissibility of using such tools in Islam. According to several Islamic law experts, the use of AI is not considered haram, yet the jihadist ecosystem is still very sceptical about its use. Nevertheless, by monitoring the users of the main IS server channel on Rocketchat, it was possible to observe how several users have reported their attempts to use AI tools to develop propaganda material without receiving any discouragement or disapproval from other users. Looking at the positive directions of the ecosystem to the above videos and the efforts of Hisad to be more accepted, it is worth asserting that, given the absence of a precise religious doctrine which prohibits its use, AI technologies could evolve into a tool for producing and diffusing propaganda material.
How to prevent a new era of pro-IS propaganda
The first step to ascertain a possible wave of AI-equipped media operators is content analysis. Once the material used has been identified, it is possible to target the supply chain of images and news. In particular, the targeting of Nashir channels, which are considered bottlenecks in the pro-IS digital ecosystem, could affect video production capacity and visibility. Secondly, AI providers, internal moderation, and access policies should impede the production of terroristic content within the platforms. The study of the propaganda released and shared by jihadist organisations, such as IS, has enabled us to identify specific recurring elements (content, visual motifs, and expressions) that can be exploited to train AIs aimed at terrorist pattern recognition. The use of AI capable of recognising and reporting terrorist patterns becomes increasingly necessary. Ali Fisher and Nico Prucha are currently engaged in the development of an advanced artificial intelligence system. This AI system is designed to be authentic and evidence-based, ensuring that its functionalities are rooted in verifiable data. Its core capability includes the recognition of the theocentric nature characteristic of the Islamic State. By launching a rapid categorisation of the material and leveraging vast datasets, the AI aims to accurately interpret and analyse the complex ideological underpinnings that define this movement. The goal of their work is to create a tool that not only enhances our understanding of such movements but also contributes to a more effective content identification within social media.
Federico Borgonovo is a research analyst at the Italian Team for Security Terroristic Issues and Managing Emergencies – ITSTIME. He specialized in digital HUMINT, OSINT/SOCMINT, and Social Network Analysis, oriented on Islamic terrorism and RWE. He focuses on monitoring terrorist networks and modelling recruitment tactics in the digital environment, with particular attention to new communication technologies implemented by terrorist organizations.
Alessandro Bolpagni is a research analyst at the Italian Team for Security Terroristic Issues and Managing Emergencies – ITSTIME. He specialised in International Relations, geopolitics in the Middle East, OSINT, Digital HUMINT, and SOCMINT. His research activities are oriented on jihadist organizations and paramilitary companies in Africa and the Middle East.
Silvano Rizieri Lucini is a research analyst at the Italian Team for Security Terroristic Issues and Managing Emergencies – ITSTIME. He specialised in digital HUMINT and OSINT/ SOCMINT oriented on Islamic terrorism, Whitejihadism, and RWE. He focuses on monitoring terrorist networks, with particular attention to new communication strategies implemented by terrorist organisations.