This weekly update covers important news about keeping AI agents safe and protecting data. AI agents are computer programs that can work on their own without a person telling them every step to take.

Companies are moving very fast to use AI agents, but they are not protecting them well enough. One report found that about two-thirds of business leaders do not understand AI agents very well. Also, six out of ten leaders said that AI threats are growing faster than their security teams can handle. This is a big problem because AI agents touch almost everything in a company's computers.

Token Security, a company in Israel and New York, started a new research team to find problems with AI agents before they cause damage. The team looks for weak spots that hackers could use to attack companies. They already found a serious problem where someone could fake passwords and steal data from Salesforce customers.

Zenity, another security company, won an award for protecting AI agents. Zenity watches how AI agents work across different platforms like business software, cloud computers, and personal devices. They use special methods to see what the agent is doing at every step.

Researchers found a new and scary way that AI agents can leak secret company information. If an attacker hides instructions in a normal webpage that an AI agent reads, the agent might follow those hidden orders. The agent could search the company's secret files and send them to a bad person's computer without anyone knowing.

Salesforce, a big software company, says it is testing its new AI agent called Agentforce very carefully. The company is making sure the AI agent acts the right way and does not do anything dangerous. They are also giving customers tools so they can control what the AI agent is allowed to do.

Palo Alto Networks, a security company, created a new tool called Cortex AgentiX to help companies build and control AI agents safely. This tool lets companies create their own AI agents and make sure the agents only do what they are supposed to do.

PwC and Stripe announced they are working together to help companies use AI agents for shopping. They created new rules to keep shopping safe when AI agents help people buy things.