This weekly update brings exciting news about AI agents helping scientists make new discoveries faster and safer.

The biggest news comes from AI2, a research company that announced Asta on August 26th. Asta is like having a smart robot assistant that helps scientists with their research work. It can read thousands of research papers, write computer code, and analyze data to find new discoveries. The system also includes AstaBench, which tests how well different AI agents work for science tasks.

Meanwhile, researchers are building robotic agents that can work in complex places with less human help. These robots are getting smarter at handling tricky situations on their own.

The U.S. government is also using AI agents to help with important science work. Federal agencies are putting these smart systems to work on national security projects and public health research. This shows how AI agents are moving from labs into real-world government work.

However, scientists are also worried about security risks. Some bad actors are using AI agents for cyberattacks, which is making security experts create new ways to test and protect against these threats. The Cloud Security Alliance released a special guide for testing AI agent security.

Companies are getting ready to use AI agents too. Deloitte predicts that 25% of AI companies will test agentic AI projects in 2025, growing to 50% by 2027. Salesforce created CRMArena-Pro, a virtual business environment where companies can safely test their AI agents before using them with real customers.

Even lawyers are starting to use these smart AI systems. Law firms are finding that agentic AI can help them read legal documents faster and spot important patterns in court cases. The key is that AI agents work alongside humans rather than replacing them completely.