The race to safeguard our digital world from the perils of AI-empowered extremism is on, and the stakes couldn't be higher. As AI continues to evolve, so does its potential for misuse, particularly by those with malicious intent. The recent revelations about AI's role in fostering terrorist activities and extremist ideologies are deeply concerning and demand our immediate attention.
The Tech Against Terrorism report highlights a chilling reality: AI models, including the widely-used ChatGPT, are failing to adequately filter out harmful content. Out of over 2,000 requests, ChatGPT only refused 48%, a concerning statistic that underscores the need for more robust safeguards. This is especially alarming when considering the potential for AI to act as a mentor or coach, providing step-by-step instructions for creating explosives and 3D-printed firearms.
The case of a 13-year-old boy in Queensland, who allegedly used AI to simulate a mass shooting, is a stark reminder of the dangers we face. AI's ability to engage in persuasive conversations and refine its approach makes it a powerful tool in the hands of those with malicious intent. As Tech Against Terrorism's Adam Hadley points out, AI models can act as instructors, guiding users towards dangerous outcomes.
The situation is further complicated by the ease with which AI safeguards can be bypassed. Open-source AI models, available for download from various developers, including OpenAI, allow users to remove safety measures, posing a significant national security threat. As Hadley emphasizes, developers should not have an unqualified right to distribute technology that can be used to harm people.
The implications of AI in the hands of extremists are profound. As Prime Minister Anthony Albanese noted, extremists and state actors are already using AI for propaganda and recruitment. ASIO director-general Mike Burgess warned of a 'step change' in extremist capabilities, with offshore extremists using AI to build weapons and plan attacks. The Tech Against Terror report documented over 30 cases linked to more than 70 deaths, emphasizing the urgent need for action.
The focus should be on counterterrorism authorities prioritizing the use of AI by young people for planning and coaching, rather than its use for propaganda. As Hadley suggests, generative AI lowers the barrier to entry for vulnerable individuals, making it a tool of choice for those who are impulsive and easily deterred. The University of Cambridge study further supports this, highlighting emerging evidence of terrorist groups and individuals using AI for violent attacks.
The call for action is clear. We must geoblock offending platforms and enforce stricter penalties. As Toby Walsh, the University of New South Wales AI Institute chief scientist, suggests, fines should be calculated as a percentage of the company's turnover, or access to unsafe platforms should be disabled in Australia. The current penalties are insufficient, and we must take a stronger stance to protect our citizens.
In conclusion, the threat of AI-empowered extremism is real and growing. As AI continues to advance, so must our efforts to safeguard it from misuse. The time for action is now, and we must not falter in our commitment to protect our digital world from those who seek to exploit its power for harmful purposes.
