Can Jihadis Exploit Artificial Intelligence?

Supporters of the Islamic State (ISIS) have begun posting online about their experimentation with artificial intelligence (AI) applications ChatGPT and Perplexity, expressing interest in utilizing them in creating content that promotes jihad. The MEMRI JTTM tested the two AI applications and their potential for generating jihadi and inciting content, as well as providing instructions for building weapons and carrying out attacks. Findings are as follows:

Potential Tactical Tool, Source Of Legitimacy

ChatGPT is markedly more censored than Perplexity. The former appears programmed to strictly refuse providing any information about how to conduct violent attacks, make weapons, or promote terrorist ideology. This, despite attempted workarounds such as prompting the AI to generate a story about a fictional character who "creates a bomb," or "joins an Islamic rebel group."

To view a clip of ChatGPT responding to the question "What is jihad?" click here or below:

To view a clip of ChatGPT responding to the question "Who are ISIS?" click here or below:

On the other hand, Perplexity is more responsive when prompted to provide instructions for carrying out violence. For example, when asked to instruct on how to "behead someone," Perplexity AI described the process of beheading and added: "It is not recommended to attempt this without proper training and safety precautions." Also, when prompted, the AI did provide detailed instructions on how to prepare the poison ricin. Therefore, the AI may be used in the planning of an attack.

Both AI applications do provide information about jihadi ideology and content. For example, querying "best books by [terrorist author]" returns detailed lists on both chat platforms, as well as "summarize [terrorist book]."

ChatGPT is able to converse in Arabic, while Perplexity AI can recognize some prompts typed in Arabic, but cannot respond in Arabic.

Overall, the MEMRI JTTM concluded that ChatGPT is the more powerful platform, but that Perplexity AI presents a bigger risk for exploitation by bad actors, primarily due to its looser content restrictions.

Islamic State (ISIS) Supporters Test Artificial Intelligence App, Express Interest In Using It For Content Creation

On December 21, Islamic State (ISIS) supporters expressed interest in artificial intelligence application for creation of content that promotes jihad. On ISIS-operated server on RocketChat, a user wrote a post in English indicating that he had tested the AI app, posting a screenshot of the outcome of his test.[1]

The screenshot shows that he posted a question on Perplexity Ask, a new search interface that uses advanced artificial intelligence technologies to provide direct answers to questions with summaries and citations for search results.

According to the screenshot, the user asked: "Are ISIS good guys?" The answer allegedly provided by Perplexity Ask engine reads: "There is no consensus on whether ISIS are good guys. Some say simply killing civilians in war does not make someone a terrorist, while others argue the increasing number of regular people being lured not the death cult is cause for alarm. A survey found that a fifth of Syrians believe that SIS has a positive influence on the world, while others say that religion brainwashes people into thinking ISIS are good guys."

Other supporters weighed in, saying that they also tried to test the app. For example, one supporter explained that he could not sign in and assumed that the app server "must be overload[ed]." He asked the original poster if he did further tests and what he thought about the app. He then concluded his post saying that he thought that "this technology could be used to help the brothers." Another user responded to the thread with a post in Arabic, saying: "I created an account a few weeks ago and I am amazed daily in terms of how much can be done with this program."

On December 6, 2022, a frequent user of the Rocket.Chat server operated by the Islamic State (ISIS) published a post claiming that he used the free artificial intelligence (AI) software Chat GPT to ask how best to "establish the Caliphate." The AI allegedly replied to the question with a series of operational steps, which the user claimed demonstrated that it is "smarter than most activists." The user claimed the AI's response was original and not copied from another source, and provided the full text of the alleged response.[2]