Online Library

In 2020, OpenAI developed GPT-3, a neural language model that is capable of sophisticated natural language generation and completion of tasks like classification, question-answering, and summarization. While OpenAI has not opensourced the model’s code or pre-trained weights at the time of writing, it has built an API to experiment with the model’s capacity. The Center on Terrorism, Extremism, and Counterterrorism (CTEC) evaluated the GPT-3 API for the risk of weaponization by extremists who may attempt to use GPT-3 or hypothetical unregulated models to amplify their ideologies and recruit to their communities.

VOX-Pol Blog post.

The United Kingdom (UK) has been at the vanguard of online censorship in democracies from the beginning of the modern internet. Since the mid-1990s the government has developed distinctive patterns of regulation – targeting intermediaries, using the bully pulpit to promote ‘voluntary’ self-regulation, and promoting automated censorship tools such as web blocking – which have been influential internationally but raise significant issues of legitimacy, transparency and accountability. This chapter examines this UK experience in light of the European Convention on Human Rights (ECHR) and EU law, arguing that in key regards current censorship practices fail to meet European standards.

The chapter builds on the existing literature in two main ways. First, it assesses emerging censorship practices in the area of terrorist material and extreme pornography. Second, it considers how recent EU legislation and ECtHR case law might constrain the freedom of the UK government and force a move towards different models of censorship.

The chapter starts by outlining the regulatory context. It then takes three case studies – Child Abuse Material (CAM), terrorist material, and pornography/extreme pornography under the Digital Economy Act 2017 – and traces how censorship has evolved from one context to the next. These systems are then evaluated against the standards set by European law and in particular Articles 6 and 10 ECHR, the Open Internet Regulation, and the Directives on Sexual Abuse of Children and on Combating Terrorism.

Recognizing that militant, non-state groups utilize social media and online platforms to reach members, sympathizers, and potential recruits, state agencies and social media corporations now increasingly regulate access to accounts affiliated with such groups. Scholars examining deplatforming efforts have, to date, focused on the extent of audience loss after account restrictions and the identification of strategies for regrouping online followers on the same or different platforms over time. Left unexplored is if and how militant non-state groups adapt their official messaging strategies in response to platform restrictions despite continuing online access to them. To begin to fill that gap, this study compares ISIS’s 550 images displayed in the group’s official newsletter al-Naba 6 months before and after Europol’s November 2019 take-down of terrorist affiliated accounts, groups, channels, and bots on Telegram. It conducts a content analysis of images related to militaries and their outcomes, non-military activities and their outcomes, and presentational forms. The findings demonstrate that ISIS visually emphasizes its standard priming approach but shifts its agenda-setting strategy. While retaining some of its standard visual framing practices, the group also alters frames, particularly those related to images showing opposing militaries and military outcome.

This paper examines the need and possibility for developing online resilience-based approaches in response to COVID-19 vaccine conspiracies, often linked to the far right. Examining three datasets collected between December 2020 and April 2021, this paper details conspiracy narratives that have developed around COVID-19 vaccines, with specific focus on understanding the deployment of the idea of a planned pandemic or so-called ‘#plandemic’. This is then used to consider where existing resilience-based approaches to countering off-line polarisation and extremism might posit an appropriate online response. The article identifies four key #plandemic framings of COVID-19 vaccines — as control, as reset, as unnecessary and as unsafe — and analyses how these themes are constructed, to find that they are often created through hostile and confrontational interaction with other users. Based on these findings, the conclusion suggests companies shift their focus away from ‘negative’ approaches to content moderation (e.g., content removal) and towards resilience-building responses that cultivate flexible individual identities, build community support networks, and/or engage users with national and supranational democratic structures, as a more effective response to the sharing of online conspiracies.