This week, the focus on legal and regulatory frameworks for agentic AI continues to build on earlier developments. The European Union's AI Act is a big deal because it's the first major law to regulate AI. It sets rules based on how risky the AI system is. The EU is also planning a voluntary code of practice by mid-2025 to help companies get ready. This is important for agentic AI, which are AI systems that can act on their own, because they are often in regulatory gray areas.

There are several key legal risks when using agentic AI. One is transparency and explainability: it can be hard to understand why the AI made a decision, especially when the law requires explanations. Another is bias and discrimination: if the AI acts in a biased way, it could break fairness laws. Privacy and data security is also a big worry because agentic AI can access a lot of personal information, and we must follow laws like the GDPR in Europe and the CCPA in California. Accountability and agency is tricky because when an AI agent makes a harmful decision, it's not clear who is to blame. Finally, agent-agent interactions (when multiple AIs work together) can lead to unexpected behavior and might bypass safety rules.

To address these challenges, some experts suggest treating personal AI agents as fiduciaries. That means they would have to act in the best interest of the user, similar to how a doctor or lawyer must act for their clients. This idea includes three parts: creating legal frameworks for fiduciary duty, encouraging tools like insurance and monitoring services, and designing agents to keep sensitive data local. Without clear rules, people might not trust AI agents with important tasks, which could slow down the adoption of this technology.