Companies are building with AI — and exposing new security risks

The utilization of AI technology is rapidly expanding, leading to the creation of additional AI-related infrastructure that is now accessible on the public internet, according to fresh data from Censys disclosed to Axios. This surge in AI development is not solely limited to the creation of new models and chatbots, but also the emergence of a vast network of software that businesses must now manage, update, and secure.

The threat level is elevated, as many of these tools serve as conduits between AI applications and sensitive corporate data, providing potential avenues for cybercriminals to infiltrate company systems. Researcher Emily Austin from Censys explained that these tools can connect to substantial, sensitive systems, which extends beyond the tool itself.

As of July, Censys identified almost 268,000 internet-facing systems utilizing 44 distinct AI products, a significant increase from approximately 164,000 systems in October 2025. These systems encompass software essential for connecting AI models to databases, corporate applications, and other services. The ease with which AI applications can be built and hosted is complicating the task of tracking and ensuring the security of the software used by employees.

Censys CEO Zakir Durumeric noted that companies are already finding it challenging to keep tabs on their existing software, even before the introduction of AI-powered applications. These new AI applications often have the capability to access sensitive company information, such as emails, financial records, and internal databases.

N8n, a workflow automation platform capable of connecting AI agents to various applications and services, was the most widely exposed product, with nearly 114,000 instances deployed on the internet, including around 77,000 vulnerable versions of the software. Additionally, over 54,000 control panels for OpenClaw, an open-source AI agent that users can link to personal accounts and other services, were found to be exposed.

While the data does not confirm whether these deployments are susceptible to exploitation, they underscore the risks associated with exposing software that can access sensitive company systems. The ultimate takeaway is that businesses must treat AI infrastructure similarly to other software they deploy, with regular reviews, updates, and security measures in place for these systems.