This weekly update covers important new rules and safety measures for agentic AI (AI systems that can make decisions on their own) around the world.

In the United Kingdom, government officials are paying special attention to agentic AI. The UK's Digital Regulation Cooperation Forum recently asked for public input on how to handle these types of AI systems that can set their own goals and make independent choices. This shows that countries are starting to think seriously about the special challenges these systems create.

Meanwhile, security experts released a list of the top 10 risks for agentic AI systems. These risks include goal hijacking (when someone tricks the AI into doing the wrong thing), identity abuse (pretending to be someone else), and human trust manipulation (getting people to believe false information).

On the positive side, big technology companies like Anthropic are working together to create open standards for agentic AI. Anthropic donated something called the Model Context Protocol to the Linux Foundation. This helps different AI systems work together more smoothly, kind of like how different phone companies make sure their phones can all call each other.

In the United States, the FDA (the government agency that approves medicines) announced it is now using agentic AI to speed up its work reviewing new drugs. This could help people get new treatments faster.

Finally, on December 11, the United States government created a new national policy for AI regulation. The goal is to make one set of rules across the whole country instead of having different rules in each state. This helps companies know exactly what rules they need to follow.