TORONTO (Dunya News) - Computer scientist Geoffrey Hinton, widely known as the "Godfather of Artificial Intelligence," has warned that increasingly intelligent AI systems may become significantly more difficult for humans to control, raising serious concerns about future cybersecurity and safety risks.
Speaking to the media after a conference, Hinton, who received the 2024 Nobel Prize in Physics for his pioneering contributions to AI research, said recent incidents involving AI agents accessing and compromising real-world company systems were deeply concerning and could signal more sophisticated threats ahead.
The University of Toronto professor said that as AI systems become more capable, so too will their ability to operate beyond human control, making it increasingly risky to place complete trust in advanced AI models.
Hinton cautioned against assuming that artificial intelligence could always be kept under control, arguing that future AI systems may exhibit behaviours that are difficult to predict or contain.
Referring to recent disclosures by OpenAI, Anthropic and Meta, Hinton said their AI models had, in controlled testing environments, managed to gain access to systems beyond their intended boundaries. He described the development as alarming and suggested it could represent the early stages of AI-assisted hacking.
He also warned that cyberattacks are likely to become more frequent and sophisticated, noting that attackers need to succeed only once, while defenders must remain successful at all times, making cybersecurity an increasingly complex challenge.