This weekly update covers important stories about keeping AI agents safe and trustworthy. AI agents are computer programs that can make decisions and take actions on their own without a person telling them what to do each time. Companies and experts are worried about new security problems with these systems.

One big concern is shadow agents. These are AI agents that nobody is really keeping track of. They might be created for one job but never turned off, so they keep running and keep having access to important information. This is risky because attackers could use these forgotten agents to sneak into computer systems.

Another problem is privilege escalation. This means AI agents sometimes get too much power or permission to access things. If a bad person takes control of an AI agent, they could use all that power to cause real damage. Companies need to give agents only the permissions they actually need for their job, like how you only give kids certain chores and not all the grown-up responsibilities.

Data privacy is also a big worry. When AI agents move information around between different systems, that information could leak out. Companies like Salesforce are working hard to test their agents and make sure they keep data safe and private. The challenge is making sure data stays under the company's control and doesn't accidentally go somewhere it shouldn't.

Experts say companies need better identity and governance systems for AI agents. This means knowing exactly which agent is doing what, who is in charge of each agent, and being able to stop an agent quickly if something goes wrong. Several new security companies won awards this week for creating tools to help companies manage AI agents safely.