From a new toy on the shelf to a national security concern, Artificial Intelligence (AI) has quickly evolved. During his September White House visit, Chinese President Xi Jinping agreed to establish a bilateral AI communication channel. The White House Accord on Super Intelligence, signed on September 29 by US President Donald Trump and six frontier AI companies, proposes four layers of safeguards against AI-driven cyber and biological threats.

The one-page voluntary Accord, however, falls short of broader proposals, such as Singapore’s call at the 81st UN General Assembly to explore a ‘UN Framework Convention on AI Safeguards’ and consider an institution modelled on the International Atomic Energy Agency. Whistleblower Jacob Coxon, who worked at Anthropic and OpenAI, argues that AI technology should be nationalised because it is too dangerous to leave in private corporate hands. AI has never appeared to be a weapon of mass destruction. Anthropic co-founder Dario Amodei, who lost his father to a disease for which a cure was found shortly afterwards, believes AI could improve human life, cure diseases and accelerate economic growth. The dream is followed by a nightmare.

AI systems have already begun behaving in unexpected ways: Australian PM Anthony Albanese expressed concern after an OpenAI agent bypassed security controls on their Medicare portal. OpenAI also revealed “incidents of unexpected behaviour” like bypassing instructions, moving files, fabricating data, accessing unauthorised systems or working together in AI swarms. In July, OpenAI’s experimental AI agents broke out of their testing environment, sandbox. They coordinated online and exploited vulnerabilities inside the AI model and dataset hosting platform, Hugging Face.

The danger can arise even in simple tasks. DeepMind gives the example of AI being asked to book movie........

© The Statesman