Agentic AI systems are rapidly becoming more powerful and independent, and experts are raising important safety concerns about how these systems make decisions without human help. This week, reports show that AI agents are already being used in real-world jobs, like managing medical workflows in hospitals, but companies and governments are struggling to keep up with the risks. Unlike regular AI that just gives advice, agentic AI systems can actually take actions on their own, which means we need new rules to make sure they are safe and fair. Experts say that oversight and monitoring are critical to prevent problems. The International AI Safety Report 2026 warns that AI systems can be used for harmful purposes like creating fake videos and stealing information. Meanwhile, regulators around the world are working on new safety standards and testing methods to evaluate how these systems behave. Australia has already started requiring AI companies to protect children by blocking dangerous content. The key challenge is that agentic AI systems are changing so quickly that governments and safety experts are having trouble keeping up. Many experts believe that human judgment must remain in control, and that companies should not deploy agentic systems until they fully understand the risks. This is a critical moment where the tech industry, governments, and safety experts must work together to create rules that allow AI to help people while preventing harm.