Palo Alto Networks Unit 42 has found a security "blind spot" in Google Cloud's Vertex AI platform This article explores security holes google. . An attacker could use this flaw to get into an organization's cloud environment and access sensitive data without permission.

Google has since changed its official documents to make it clear how VertexAI uses accounts, resources, and agents. The big tech company has also told customers to use Bring Your Own Service Account (BYOSA) to replace the default service agent and follow the principle of least privilege (PoLP) to make sure the agent only has the permissions it needs to do its job.

"By default, giving agents too many permissions goes against the principle of least privilege and is a dangerous security hole by design," said Shaty ShatY of Palo Alto Networks in a blog post about the security hole. "This level of access is a big security risk because it turns the AI agent from a useful tool into a possible insider threat," said Shatay of Unit 42. He went on to say, "An attacker could use this unintentional visibility to map Google's internal software supply chain, find old or vulnerable images, and plan more attacks."

People have called the security hole a "significant security risk" for businesses that use the AI platform to make and test new products and services.

To learn more, check out Palo Alto Network's blog post about the security holes in Google's AI platform. You can also read the whole post in the "Security in the Cloud" section of the blog. You can also read the blog post on the Google Cloud Security Blog.

For more information on the same topic, check out the Security in the Cloud section of this article. You can get private help by calling the Samaritans at 08457 90 90 90, going to a local Samaritans branch, or visiting www.samaritans.org. If you need help in the U.S., call the National Suicide Prevention Lifeline at 1-800-273-8255.