Artificial intelligence is entering a new phase as researchers and technology companies move beyond traditional chatbots and digital assistants towards systems capable of independently planning and carrying out complex tasks. Known as AI agents, these systems can use digital tools, browse the internet, write and operate software, and complete sequences of tasks with limited human supervision. The International AI Safety Report 2026, prepared with contributions from more than 100 AI experts, notes that AI agents have become increasingly capable, particularly in software engineering and other computer-based activities. At the same time, the report highlights continuing limitations, with systems more likely to make mistakes when confronted with lengthy, unusual or complicated tasks.

 

The next transformation could take AI beyond computer screens and into the physical world. Researchers are developing vision-language-action systems that enable robots to understand spoken instructions, interpret their surroundings and perform physical activities. While some systems can already complete relatively simple tasks, unexpected circumstances and unfamiliar objects remain significant challenges. Energy consumption is emerging as another major consideration. The International Energy Agency projects that electricity consumption by data centres could reach around 945 terawatt-hours by 2030, with the expansion of AI expected to be a major contributor to rising demand.

 

Looking ahead, current trends suggest that AI systems could potentially perform well-defined software tasks lasting several days by 2030. However, experts emphasise that technological progress remains uncertain, with future capabilities depending on advances in AI models, computing infrastructure, robotics, safety measures and energy availability. The emerging AI era could therefore be defined not simply by machines that answer questions, but increasingly by systems that plan, act and interact with the world.