Beyond Extinction: Counterterrorism Experts Warn of Immediate AI-Enabled Radicalization

Key points

• NCITE Center researchers warn that 'abliterated' AI models, which have ethical guardrails removed, are being used by terrorist groups to plan attacks and radicalize individuals.
• Over 12,000 uncensored AI models are now publicly available on platforms like Hugging Face and GitHub, with some having over a million downloads.
• A report by AI 'godfathers' Geoffrey Hinton and Yoshua Bengio warns of an 'intelligence explosion' where AI automates its own research and development, potentially leading to superhuman capabilities within months.
• Anthropic and OpenAI have agreed to independent evaluations of their models, acknowledging that AI now produces a significant portion of their own code.
• Public concern has surged, with 64% of Americans believing AI poses a serious threat to humanity, leading to legislative proposals to ban 'superintelligence' and calls for stricter regulation.

Background

Recent developments in AI governance have been marked by a tension between rapid development and safety concerns. In September 2026, President Trump announced plans to form an 'AI Force' to accelerate U.S. AI growth, dismissing existential risks as a hoax. Concurrently, the U.S. and China engaged in discussions regarding shared AI risks and safety standards, amid accusations of Chinese firms distilling American AI models. Earlier coverage highlighted the debate over whether AI poses a real existential threat or if risks are manageable through regulation and cautious engagement.

How outlets are covering it

The primary source, Lawfare, emphasizes the immediate, tangible threat of abliterated AI models being used by terrorist groups for radicalization and attack planning, citing specific examples like Houthi-linked groups using Claude code for missile development. In contrast, The Guardian focuses on the theoretical risk of an 'intelligence explosion' where AI automates its own R&D, potentially leading to a loss of human control. CNN highlights the emotional and political impact of these fears on the public, particularly parents, and the resulting pressure for legislative action. While Lawfare and The Guardian both warn of severe risks, Lawfare focuses on current, human-driven misuse of uncensored AI, whereas The Guardian and CNN focus on the future potential of autonomous AI systems to outpace human oversight. The Telegraph source was inaccessible due to a security block, so its perspective could not be synthesized.

Why it matters

The convergence of immediate security threats from uncensored AI and long-term existential risks from automated AI development underscores the need for robust regulatory frameworks. The availability of abliterated models on public platforms poses a direct risk to public safety, while the potential for an intelligence explosion could disrupt global power dynamics and safety measures. Public and political pressure is mounting for governments to implement transparency, independent audits, and potentially pause certain AI developments to mitigate these risks.

What to watch

Governments are expected to consider policies requiring transparent progress reports on AI R&D, embedding independent auditors in AI companies, and potentially limiting the speed of AI self-improvement. Legislative efforts, such as the proposed ban on 'superintelligence' by Senators Bernie Sanders and Rep. Greg Casar, may gain traction as public concern grows. AI companies like Anthropic and OpenAI are likely to continue engaging with independent evaluators to assess model safety, while counterterrorism agencies will monitor the use of abliterated models by extremist groups.