What is Agentic AI and Why Does It Need Safety Rules?

Agentic AI means AI systems that can make decisions and take actions on their own. Unlike a calculator that only does what you tell it to do, agentic AI can learn and choose what to do next. This is powerful, but it can also be risky. That's why safety and ethics are so important right now.

Countries Around the World Are Making New Rules

Many countries are realizing that AI is growing so fast that old rules don't work anymore. The European Union passed a big new law called the EU AI Act in 2024 to control how AI is used. Scientists who study AI safety say governments have moved from just asking companies to be careful to making actual laws that companies must follow.

Safety Problems for AI Systems in the Real World

One big problem is that AI systems are starting to move around and work in the real world. Self-driving cars, delivery robots, and medical devices all use AI. These systems need to be super safe because they can hurt people. But right now, there are more than 30 different safety rules from different groups, and they don't all work together. This makes it hard for companies to know which rules to follow.

California Moves Forward with Self-Driving Trucks

In the United States, the state of California just made new rules for self-driving trucks. Companies that want to test and use trucks without drivers must now follow these rules. The trucks need to drive at least 500,000 miles with a person watching before they can drive alone. The trucks also need to follow all traffic laws, or the police can give the company a notice saying the truck broke the rules. In an emergency, local leaders can tell all self-driving cars in an area to leave right away.

How Robots Should Make Choices

Scientists are working on how to teach AI systems to make good choices when they face hard situations. For example, if a self-driving car must choose between hitting one car or another car, which should it do? Researchers say robots need clear rules that help them pick the least harmful choice. This is very hard to figure out, and it shows why we need safety experts to work with AI builders.

Concerns About AI and the Military

Some people are worried about using AI in war and weapons. Workers at the company Google sent a letter to their bosses asking them to stop helping the military use AI in fighting. This shows that many people think it's important to talk about how AI should and should not be used. The worry is that if AI systems make weapons decisions without a person saying yes, it could be very dangerous.

Africa's Special Concerns

In Africa, leaders are saying that when other countries and companies make rules about AI, they need to think about how it affects Africa. Africa only has about 2% of the data storage centers in the world, but companies often use African workers to label data or test AI systems. Africa's leaders say that each country needs to make its own rules too, so that African people have real power to control AI in their countries.

What Experts Say We Need

Safety experts say that AI systems need to have safety checks at many different levels. The computer chips need to be safe, the software needs to be safe, and the way the AI system works in the real world needs to be safe. They also say that the people controlling self-driving vehicles from far away need special training. As AI systems get more powerful and independent, we need to keep learning how to keep them safe for everyone.