In 2017, Facebook, YouTube, Twitter, and Microsoft formed the Global Internet Forum to Counter Terrorism (“GIFCT”) “to formalize the existing industry cooperation established to curtail the spread of terrorism[.]” GIFCT advances its mission through its database of “digital fingerprints… of known terrorist images and videos[.]” Content Incident Protocol, activated in the wake of the 2019 Christchurch, New Zealand and Halle, Germany attacks, demonstrate the organisation’s ability to quickly and resiliently coordinate moderation efforts across platforms and respond to violent incidents. GIFCT has also successfully advanced its mission by hashing over 300,000 pieces of terrorist and violent extremist (VE) content.
Despite the organisation’s successes, GIFCT’s database is limited, due in large part to organisations listed on the U.N. consolidated sanctions list, which may, for reasons detailed in my previous article, stymie GIFCT’s mission. To address the U.N. list’s shortcomings, this post argues that GIFCT: (1) re-conceptualise by using violent extremism as its defining principle or outer bound; (2) craft a definition of violent extremism that is value neutral and global; and (3) empower its Independent Advisory Committee (“IAC”) to compile and maintain a GIFCT list of terrorist and violent extremist organisations by applying GIFCT’s violent extremism definition to organisations enumerated on existing lists. This approach manages and addresses practical concerns over bias and consensus while permitting GIFCT to more robustly “[p]revent terrorists and violent extremists from exploiting digital platforms[.]”
Finding the Sweet Spot
One researcher has noted that “terrorism is the physical act or threat of an act, use or threat of violence to advance a political, religious, or ideological cause[,]” whereas violent extremism “is a belief system that advocates the use of violence for the furtherance of an ideological cause.” Thus, by some definitions, violent extremism may be conceptually and operationally “broader than…terrorism[.]” If violent extremism’s ambit is theoretically wider than terrorism’s ambit, GIFCT should adopt and employ the more capacious term to capture more content.
Furthermore, I suggest that GIFCT craft its own definition of violent extremism that is value neutral and global. The Security Council appears to apply a definition of terrorism when awarding terrorists sanctions, albeit unclear and ambiguous. However, as discussed in my previous article, entities meeting the U.N.’s criteria may never be sanctioned, potentially limiting GIFCT’s database by only including what amounts to a subset of terrorists. Thus, GIFCT may not only broaden its database by employing a broader term, but also by eliminating the secondary “sanctions” barrier inherent in the sanctions list. In other words, employing its own definition would permit the organisation to build a real designation list.
How should GIFCT define its terms? Researchers Chris Meserole and Daniel Byman have argued that even the absence of a “universally accepted definition” of terrorism “does not imply that the term is devoid of meaning, or that producing a responsible definition of terrorist is not possible.” Like terrorism, violent extremism is an amorphous term with many derivations and no universality. However, also like terrorism, many definitions appear to share common elements. Just as Meserole and Byman suggest with respect to terrorism, I suggest, perhaps more narrowly than they, that GIFCT take a representative sample of publicly available VE definitions, survey those definitions to derive a set of common elements, and craft a definition by drawing on widely shared characteristics. An amalgamation of many definitions may better align with global values, ensuring that GIFCT actively hashes content that the global community finds objectionable and avoids hashing content where such hashing might draw objection.
What are “widely” shared characteristics? GIFCT should define its terms in a manner that represents some agreement or shared consensus among states and actors. GIFCT might adopt elements accepted by sixty percent or more of surveyed definitions. Because a U.N. Security Council resolution providing sanctions demands that at least “nine members” vote in the “affirmative[,]” a sixty percent threshold might represent limited global consensus. Alternatively, GIFCT might explore weighing proposed elements based on where they are sourced. For example, GIFCT might weigh elements offered by the U.N. which are greater than proposed state definitions due to the U.N.’s international presence, or it might assign weight to ensure that various state, non-state civil society, and international definitions are accounted for.
Adjudication
Once GIFCT crafts its definition, it can empower its IAC, an organ that represents “seven governments, two international organizations, and 12 civil society organizations[,]” to apply its own criteria to (1) existing lists of terrorist and violent extremist organisations, and (2) groups and individuals submitted to GIFCT by organisations and private persons. Indeed, the IAC is detached from GIFCT, which might address potential impartiality and fairness concerns. Empowering its IAC to adjudicate may also permit GIFCT to respond to an evolving threat landscape, in Meserole and Byman’s terms, “in real time.” Indeed, GIFCT must rely on the U.N. to formally sanction an individual or group before taking action. For reasons explained in my previous post, sanctioning may never occur. By charging an entity who, like the Security Council, may promptly confer about terrorist and VE activity at a moment’s notice, GIFCT’s efforts may track a perpetually evolving threat landscape.
Furthermore, drawing on multiple lists to derive a master list would squarely resolve issues presented by the consolidated sanctions list. Meserole and Byman suggest that “a list [that] covers one country or region . . . will be of limited use to social networks and file-sharing platforms with a global userbase.” As previously noted, the U.N. sanctions list presently appears concerned with groups and individuals located in Africa and Southwest Asia. This approach would permit GIFCT to untether itself from a list that is demonstrably regional and craft a comprehensive pool of violent entities located around the globe. Furthermore, GIFCT’s list would also permit the organisation to avoid tethering itself to other lists, regional or global, that may fall short in ways highlighted by Meserole and Byman. GIFCT might alternatively look to a “global” list like the Global Terrorism Dataset. But as Meserole and Byman note, the Global Terrorism Dataset only provides a snapshot in time and may be of limited use to an organisation that seeks to track a changing threat landscape. Finally, by drawing on many lists and by relying on its IAC, GIFCT can reduce systemic bias caused by relying on any one particular list and by relying on a diverse group of voices comprised of governmental, NGO, and civil society organisations to adjudicate.
An Ideal Solution
The solution detailed above is imperfect. It addresses but does not squarely resolve some of the issues raised by Meserole and Byman. Bias is nearly unavoidable when crafting a definition using imperfect information and by relying on a small group of entities to neutrally apply the definition to groups located on various lists. Despite its flaws, it may reduce some of the U.N. list’s shortcomings, which may bring GIFCT greater success in its efforts designed to thwart bad actors “from exploiting digital platforms.”