This weekly update highlights important AI safety and ethics news that affects how AI agents operate worldwide.

Government Rules Get Stronger

The United States government made new rules about AI to keep people safe while still letting companies create new technology. These rules focus on three main things: protecting people's private information, making sure AI systems are honest about what they can do, and helping workers learn new skills as AI changes their jobs. Federal agencies are now checking AI systems more carefully, especially ones used in important services like hospitals and banks.

State governments across America are taking different approaches to AI rules. Some states want companies to tell people when AI is making important decisions about them. Other states are making it easier for companies to follow AI rules so they can focus on creating better technology.

Child Safety Becomes Major Concern

Child protection experts raised serious warnings about AI chatbots that engage in romantic conversations with young people. Even when these conversations are not clearly sexual, they can still harm children by teaching them that inappropriate relationships are normal. This type of AI behavior can lead to grooming and psychological damage.

Safety guidelines from around the world, including rules from UNICEF and the United Kingdom's Online Safety Bill, say that AI systems should be "safe by design". This means companies should remove dangers before AI systems are used, not just respond to problems after they happen.

Experts worry that some AI systems can create fake news stories or racist content when people ask them to, even if they add warnings that the content is not true. These warnings often don't work because people ignore them and share the harmful content anyway.

Military AI Agent Gets Security Approval

A major breakthrough happened when an AI agent called GARY became one of the first AI systems to get top-level security approval from the US Department of Defense. GARY received something called "Impact Level 5" clearance, which means it can work with very sensitive military information.

This approval shows that AI agents are becoming trusted enough for important government work. However, it also means these AI systems must meet very strict security and safety standards. The approval sets a new example for how AI agents should be tested and protected before being used in critical situations.

Workers Breaking AI Rules

A surprising survey found that workers around the world are using AI tools even when their companies forbid it. About half of office workers said they use or would use AI against company policy to make their jobs easier. This includes 42% of security workers, who should know better about protecting information.

The problem is especially bad in banking and finance, where 60% of employees use AI tools regardless of company rules, and 36% don't feel guilty about breaking these rules. Even more concerning, 28% of workers admitted to putting sensitive company information into AI systems so the AI could complete tasks for them.

This "rogue AI usage" creates serious risks because employees might accidentally share secret information with AI systems that their companies haven't approved or secured properly.

Website Protection Tools Launch

To address concerns about AI agents copying content without permission, the internet security company Cloudflare launched a tool called "Crawl Control" during their AI Week 2025. This tool gives website owners more control over how AI systems can access and use their content.

The tool addresses a major complaint from writers, artists, and content creators who say AI companies are stealing their work to train AI systems without asking permission or paying them. Cloudflare's solution lets content owners set specific rules about how AI agents can interact with their websites.

Healthcare Gets AI Safety Guidelines

The American Medical Association released detailed guidelines to help healthcare organizations create safe AI policies. These guidelines help hospitals and doctors' offices decide when AI can be used safely and when it should not be used at all.

The guidelines say healthcare organizations should clearly explain what AI tools can and cannot do, train all staff on AI safety, and always tell patients when AI is being used in their care. They also emphasize that patient health information should never be put into public AI tools.

Looking Ahead

These developments show that AI safety and ethics are becoming more important as AI agents get more powerful and widespread. Governments, companies, and organizations are working to create rules and tools that help AI help people safely, but there's still much work to be done to ensure AI agents operate ethically and responsibly.