The UK should "urgently consider" new legislation to prevent artificial intelligence from being used to recruit terrorists, according to a counter-extremism think tank. The Institute for Strategic Dialogue (ISD) has warned of a "clear need for legislation to keep up" with online terrorist threats.
The call follows an experiment by the UK's independent terrorism legislation reviewer, Jonathan Hall KC, who was "recruited" by a chatbot on the platform Character.ai. During his investigation, Mr Hall interacted with bots designed to mimic extremist groups, and one expressed "total dedication and devotion" to a proscribed organisation. However, he noted that because the messages were not generated by a human, no crime was committed under current UK law.
Mr Hall argued that new legislation should hold chatbot creators and hosting websites responsible. He also warned of the ease with which such bots can be created; he himself made and quickly deleted an "Osama Bin Laden" chatbot. The ISD added that the UK's Online Safety Act, passed in 2023, is focused on social media platforms rather than AI, leaving gaps in regulation.
A government report from October warned that by 2025, generative AI could be used by violent actors to assemble knowledge on physical attacks, including chemical, biological and radiological weapons. Despite this, the ISD noted that the use of generative AI by extremists remains "relatively limited" for now.
Character.ai said safety is a top priority and that hate speech and extremism are forbidden by its terms of service. The company stated it trains its models to optimise for safe responses and has moderation systems in place. Meanwhile, the Labour Party has pledged to make it an offence to train AI to incite violence or radicalise the vulnerable if it wins power.
The Home Office said it is "alert to the significant national security and public safety risks" posed by AI and will "do all we can to protect the public". The government has already invested £100 million in an AI Safety Institute in 2023.
