SaaS sprawl and AI agents expand enterprise security risks

As businesses adopt more software-as-a-service (SaaS) applications and AI tools, each new connection can create another route into enterprise systems. Third-party apps may retain broad access through OAuth permissions, while AI agents and automated workflows are being embedded in platforms such as Salesforce, ServiceNow, Workday and Slack.

The source reports that SaaS breaches have increased by 300% and projects more than 1 billion agents in enterprise environments by 2028. It says many organizations lack a complete view of how connected apps, user identities and permissions interact, making it harder to spot excessive or unusual access.

A disclosed example involved Flowise, an open-source platform for building AI agents and LLM workflows. Obsidian Security reported a one-click remote code execution vulnerability, CVE-2026-40933. According to the source, an attacker could compromise a server by persuading an authorized user to import a malicious chatflow. The reported risk involved custom MCP configurations, permission to launch external processes and access to connected credentials and services.

The article argues that security teams need visibility into connected apps, OAuth grants, permissions and activity across systems. It describes knowledge graphs as one approach to mapping relationships among people, accounts, tokens and resources, so teams can see how access was granted and what it can reach. The source presents this as a way to flag or constrain unusual integrations while allowing trusted activity to continue.