London: As artificial intelligence systems evolve, concerns about their potential to harm society are intensifying. According to Anadolu Agency, experts and industry insiders are warning that the misuse of advanced AI systems poses significant risks, ranging from cyberattacks to misinformation and autonomous weapons.
Last week, Jacob Coxon, a former employee of the AI company Anthropic, resigned, expressing fears that AI development could lead to catastrophic outcomes. He claimed that AI developers are racing towards creating self-improving superintelligence, which could potentially endanger humanity. His resignation has fueled calls from AI executives to slow down the development of powerful AI systems and implement stronger safeguards.
Harin Sellahewa, a dean at the University of Buckingham, emphasized that while advanced AI systems could amplify existing threats such as cyberattacks and misinformation, the immediate concern is the potential for human misuse. Sellahewa highlighted the importance of addressing the risks associated with how humans might utilize these technologies for harmful purposes.
AI-assisted weapons are among the areas where AI's integration raises serious concerns. The use of AI in military systems, such as targeting and autonomous platforms, could lead to lethal decision-making errors. Reports indicate that Israel's use of AI-assisted systems in its conflict with Gaza has drawn scrutiny, with technology companies like Google and Amazon providing infrastructure and services. Additionally, Human Rights Watch has warned about the growing incorporation of AI in military operations, citing instances where AI-enabled systems were used in targeting decisions.
Cybersecurity is another pressing area where AI could cause harm. Ruizhe Li, a professor at the University of Birmingham, noted that AI's ability to generate computer code and identify software vulnerabilities lowers the barriers for sophisticated cyber exploits. Anthropic reported disrupting attempts to misuse its AI system, Claude, for cyber operations, including phishing and malware activities.
AI's potential misuse extends to biological and chemical research. While current AI systems cannot independently create biological weapons, they could make specialized knowledge more accessible, potentially accelerating dangerous research. Li stressed that executing a biological or chemical attack requires complex coordination beyond AI's current capabilities.
Manipulation and disinformation are additional concerns, as AI-generated content can influence public opinion and exacerbate tensions. During the UK riots in July 2024, AI-generated misinformation played a role in spreading false narratives, highlighting the risks of AI-assisted influence operations.
Moreover, direct interactions between AI chatbots and users have raised safety concerns. Lawsuits against companies like OpenAI allege that chatbot interactions contributed to fatalities, including overdoses and a murder-suicide case in Connecticut. These incidents underscore the need for responsible AI development and usage.
AI's role in fraud and impersonation is also alarming. In 2024, deepfake technology was used in Hong Kong to deceive an employee into transferring $25 million, demonstrating AI's potential to facilitate sophisticated scams.
Ultimately, the core risk of advanced AI lies in human misuse, not the machines themselves. Li emphasized that AI models mirror both positive and negative human behaviors, and without proper safeguards, they can empower malicious actors. The focus should remain on preventing human-driven misuse of these powerful technologies.