Background
In the summer of 2024, Gaza solidarity encampments were established worldwide on university campuses. Beginning in the US, media attention was focused sharply on the actions and demands of students. In Canada, the McGill University site was among the most discussed by traditional and social media platforms. This was likely due to McGill’s reputation and the protesters' success in delaying the school’s efforts to dismantle the camp. The McGill encampment lasted from April 27, 2024 to July 10, 2024, when it was demolished by a private security firm hired by McGill’s administration and aided by Montreal police.
Across Canada, conspiracy theories about the encampment, namely accusing its participants and supporters of being funded by Iran, were rampant. The National Post and other PostMedia outlets regularly reported that the McGill University encampment was supported, in part, by Iran as part of a foreign interference campaign. In traditional media, talk radio and opinion columns were essential in promoting these narratives. Talk radio and op-eds are not subject to typical journalist standards and can perpetuate baseless claims without meaningful professional consequences.
While reports of Iranian interference in Canadian protests were based on little other than hearsay, a recent report by XPOZ — an Israeli cyber-security company that utilizes “advanced generative AI to fight against digital misinformation” to protect organizations, companies, and states from “hostile narrative attacks” — has given new legs to these theories. Proponents of these conspiracies are now using an XPOZ report about Iran interference to bolster their claims. However, XPOZ is a highly partisan source with suspicious and opaque research methods. As a self-described social tech company, XPOZ claims to be a civilian, non-political enterprise. However, it's also one that has direct ties to the Israeli occupation forces as it was founded by technology and cyber expert veterans of the Israeli Defence Technology Elite units.
While XPOZ claims to have uncovered a significant disinformation campaign with potential ties to Iran, several methodological and contextual issues raise concerns about the validity and reliability of their findings. We strongly recommend that journalists, columnists, and editors refrain from citing XPOZ due to these serious issues.
The timing of the report is notable. Earlier this summer reports about an Israeli campaign to influence Canadian politics on social media. The reputable Digital Forensic Research Lab (DFRLab), run by the Atlantic Council, unveiled an Israeli influence campaign in Canada in their report, “Online tool helps social media accounts amplify pro-Israel messages.” Global Affairs Canada confirmed elements of the Israeli social media campaign in Canada. It is notable that not long after a well-documented Israeli influence campaign in Canada and the US targeting the public as well as politicians, a dubious report would emerge from XPOZ accusing Iran of similar activities.
It is important to note that, at the time of publication, the XPOZ report on Iranian interference and support for McGill’s Gaza solidarity encampment is not available publicly. Their report has only been provided to specific people, including Warren Kinsella and Ari David Blaff, staunchly pro-Israel writers for PostMedia. Reference to the XPOZ material comes from its reporting in Iran International by Negar Mojtahedi, who published the alleged claims by XPOZ in an article sensationally titled, “Iran masterminded anti-Israel protest in Canadian university.” Further information on XPOZ was collected through their previous public statements online and through video conferences. Through a short nine-page report, they selectively highlighted examples of foreign interference by Iran and Russia at the Brandeis Institute. As seen in the Brandeis video, the report excerpts included a claim about Iran's social media influence on the McGill encampment.
Negar Mojtahedi, a journalist for Iran International, and Irad Ben-Gal, an Israeli Professor focused on Machine Learning and one of the Co-Founder of XPOZ, were contacted for clarification and access to the report, but neither replied before the publication of this essay.
Furthermore, Canadian journalists have been attempting to access the report, but seemingly, XPOZ refuses to provide a copy to the public. Toronto Star columnist Justin Ling stated in a recent article that he asked the company repeatedly, for a copy of the report. However, an XPOZ representative refused to hand it over or provide any detailed information about its methodology and how many tweets were sent out by Iran. The only response by XPOZ to Ling was that they “scraped publicly available tweets, made their best guess about where the accounts were based, and then used their ‘proprietary model to establish inauthenticity.’
Because they refuse to share their report, no journalist can either confirm or validate their claims.
Lack of Data Transparency
One of the primary concerns with the XPOZ report is the lack of transparency regarding the data and methodologies used. Without access to the raw data or detailed explanations of the algorithms employed, it is impossible for independent researchers, journalists, or even the public at large to verify the claims they have made. One of the fundamental principles of scientific research is transparency, allowing others to review and replicate findings. XPOZ does not provide access to the raw data or detailed methodologies to derive its conclusions. This lack of transparency makes it impossible to validate their claims. Selective screenshots in media articles don’t count.
Given that the data has only been provided, as far as we know, to a few highly partisan journalists, this makes the reason for skepticism even more significant.
Additionally, for this type of report, robust research typically undergoes peer review and academic scrutiny, ensuring that methodologies are sound and conclusions are reliable. XPOZ's findings do not appear to have been subjected to any review processes, which undermines the credibility of the conclusions drawn by XPOZ.
Potential Bias and Conflict of Interest
As previously mentioned, the founders of XPOZ, including Prof. Irad Ben-Gal, have strong ties to Israeli military and intelligence services.
Prof. Irad Ben-Gal is an academic with a strong background in AI and machine learning, having held significant positions and conducted extensive research. He also has an Israeli military background, including serving as an intelligence officer. The other co-founders, Ori Shaashua and Ran Farhi, also have similar credentials, with backgrounds in AI and intelligence in the Israeli military, further linking the organization to Israeli military and intelligence operations.
While their expertise in AI and cybersecurity is undeniable, their Israeli military connections also introduce a significant anti-Arab and anti-Palestinian bias in their interpretation of data given their direct vested interest in Israeli national security interests.
In a politically charged context such as the current siege on Gaza by the Israeli military, the potential for their pro-Israeli bias calls into question the objectivity of the analysis given their background and findings reflected the agenda of the Israeli state.
XPOZ executives, in various statements online, have expressed a clear focus on combating anti-Israel ‘disinformation’ to protect the state’s image. This mission is far from objective, making it difficult to believe their claims of “anti-Israel” dissent or criticism of Israel as inauthentic or manipulated by foreign entities.
For instance, in their Brandeis Institute presentation, they claim that the use of prominent hashtags like #FreePalestine or #Rivertothesea are statements from bots that promote inauthentic narratives. These slogans are quite popular in the pro-Palestinian advocacy space, and labeling them as manipulated narratives also serves to delegitimize genuine grassroots movements that use these chants.
It is also unclear if XPOZ funded this research or was contracted to do so. This lack of transparency is worrying as it is possible the research was conducted at the behest of an organization seeking outcomes to serve a partisan goal.
Selection Bias in Social Media Analysis
XPOZ reports a high percentage of anti-Israel posts as inauthentic compared to pro-Israel posts. This disparity may result from selection bias, where the criteria for identifying inauthentic behavior disproportionately affect certain groups. The methodology for distinguishing between authentic and inauthentic accounts is not clearly defined or disclosed.
According to the National Post, “60 per cent of anti-Israel social media posts concerning McGill were posted by fake accounts. By comparison, the company found 75 per cent of commentators critical of the encampment were authentic users. The dataset drew on nearly 150,000 posts on X, over half a million likes on the platform, and more than 65,000 comments.
CJPME finds this data unclear. While the report's claim that 60 per cent of anti-Israel posts were made by fake accounts, 75 per cent of the anti-encampment (presumably pro-Israel) posts were flagged as authentic. This raises questions about the criteria used to determine authenticity of an X account. The lack of clarity on how these distinctions were made suggests a possible selection bias that disproportionately targets certain viewpoints.
The report allegedly highlights a significant portion of Farsi and Arabic commentators supporting anti-Israel protests. This observation, while potentially valid, could also be a reflection of the demographic engaged in these discussions rather than evidence of foreign manipulation. Given that XPOZ claims that the majority of the posts were in Arabic and Farsi, this could result in an unfair representation of genuine political dissent as inauthentic or manipulated. Given the context, goal, and community drawn at the university encampment, it is quite normal for online users to write in those languages. It is also strategically unclear why a campaign of Farsi bots on social media would significantly impact Canadian discourse surrounding university encampments.
Algorithmic Limitations and False Positives
AI and machine learning tools, while powerful, are not infallible. The high percentage of posts labeled as inauthentic by XPOZ suggests possible issues with the precision of the algorithms used. False positives can occur, where legitimate posts are incorrectly identified as inauthentic, leading to a distorted understanding of the social media landscape. Furthermore, there is no indication that these data sets were peer-reviewed or analyzed afterward by a reliable third party or vetted by a human. Therefore, it would be beneficial if the information in the report was made publicly available for analysis by members of Canadian civil society interested in the information reported.
Broader Media and Geopolitical Context
The media's portrayal of the XPOZ findings lacks critical scrutiny and fails to provide a balanced view of the situation. The main source citing XPOZ is Iran International, which has known ties to Saudi funding and a clear opposition to the Iranian regime. Given their editorial stance, it is not far-fetched to believe that their framing of the narrative is in bad faith and in opposition to the plight of the Palestinians. Ms. Mojtahedi, the writer whose name is attached to the byline of the Iran International piece, has published a number of anti-Iranian regime articles, conflates protests in support of the Palestinian people in Canada as “Pro-Hamas” supporters, and essentially frames them as terrorist sympathizers.
Additionally, the piece by the National Post by Mr. Blaff, who has a documented record of publishing anti-Palestinian articles defaming Palestinians and essentially excluding or pressuring others to exclude Palestinian perspectives, as well as defaming Palestinians and their allies as antisemites for being inherently antisemitic for criticizing Israel.
This context is crucial for readers to understand the broader geopolitical influences at play and to approach the findings by XPOZ with necessary skepticism.
To conclude, while the XPOZ report on the McGill encampment provides a concerning perspective on a possible social media influence campaign by the Islamic Republic regime of Iran, the significant methodological and contextual issues raise serious questions about its reliability.
CJPME advises that Canadian media exercise greater caution and critical analysis when reporting on XPOZ's current findings. In light of these concerns, journalists and media outlets are encouraged to seek further verification and provide a more nuanced discussion of the potential biases and limitations inherent in the XPOZ report.
[Last Updated August 16, 2024]