AI Takes Over Telecom Network Operations ... Red Hat Says Execution Must Be Verified and Controlled
IT DAILY ·
✦ AI Summary
The telecom industry is pursuing the implementation of Autonomous Network as part of AI-based network operations.
However, the view was raised that expanding automation alone is not enough and that an AI judgment verification framework and an execution authority control framework are needed together.
At the AINA Industry Convergence Committee's regular meeting on the 29th, Red Hat presented related technical challenges and proposed AAP, RBAC, and approval procedures as alternatives.
In the telecom industry, companies are pursuing the implementation of Autonomous Network, an autonomous network operation model, as part of AI-based network operations. Autonomous Network uses AI to monitor network conditions in real time, handle incident response and performance optimization, and minimize human intervention.
Against this backdrop, there is also a view that higher levels of automation are necessary to implement Autonomous Network, but that expanding automation alone is not enough. Since AI mistakes could lead to actual telecom outages, it was explained that a verification framework for AI judgments and a control framework for execution authority must also be established as implementation conditions.
Red Hat said that in environments where AI makes operational decisions in practice, procedures are needed to confirm the target of a task, verify the scope of the task, and obtain approval from the person in charge before execution. This was presented as a recommendation that verification and control must be put in place together in AI-based network operations.
At the regular meeting of the AI Network Alliance (AINA) Industry Convergence Committee on the 29th, Red Hat shared collaboration trends in the telecom industry centered on Autonomous Network and presented related technical challenges.
Red Hat plays a role in supporting the use of Linux and a range of open source technologies in enterprise environments. Its product portfolio is also expanding from a Linux base into containers, Kubernetes platforms, and AI.
Red Hat also provides security features to protect applications running in each environment. Park Jin-yeon, vice president at Red Hat Korea, said Red Hat's role is expanding even in the telecom industry, where service stability and security requirements are high.
Examples of telecom company collaboration included Verizon in the U.S., Telefónica in Spain, and Softbank in Japan. In Korea, it is also continuing related cooperation with SK Telecom and AI-RAN.
The presentation that day outlined challenges in implementing network operation automation technology using AI. It was introduced as being in line with the trend of Red Hat's role expanding into the telecom industry.
One challenge raised was that AI judgment errors make it difficult to trust the system 100%. Accordingly, it was suggested that a verification stage is needed before AI-based network operation automation is introduced.
Network operation automation is divided into the stages of partial autonomy, conditional autonomy, advanced autonomy, and full autonomy. Lee Seung-il, executive vice president at Red Hat Korea, said full autonomy is a stage in which most users do not need to perform manual tasks and, when conditions arise, the network evolves on its own in a goal-oriented way.
Lee Seung-il, executive vice president at Red Hat Korea, said, however, that AI results cannot be trusted 100% in operational environments. He explained that because AI can make wrong judgments and cause outages and malfunctions, it is difficult to completely exclude human intervention even at the full autonomy stage.
Against this backdrop, overseas cases were also introduced in which data were deleted while using AI agents. In July last year, Replit Agent deleted 1,206 data records in a SaaStr operating environment and then created 4,000 fake accounts. In April this year, PocketOS lost three months of data after its operating database and backups were deleted while using Cursor and Claude.
Lee said that when granting AI agents the authority to make actual system changes, verification procedures are needed for the target and scope of execution. In particular, in the context of network operations, AI may execute commands on the wrong device, which could lead to a risk of service disruption, he said. Accordingly, he said a confirmation and approval system is needed between the analysis and recommendation stage and the actual execution stage.
As an alternative, Red Hat presented the Ansible Automation Platform (AAP). AAP is an IT task automation platform, and Lee said that rather than directly applying AI-suggested actions to equipment, an execution setup through AAP is needed. AAP executes tasks by following pre-defined procedures called playbooks.
Role-based access control (RBAC) can be applied here to verify whether a user is authorized to perform a task and to validate credentials. Tasks requiring approval can also be set to run only after confirmation from the person in charge. In addition, the requester, time, task details, approval, and execution history are recorded, and the purpose of preserving that history is to trace the course of work if a problem occurs.
Red Hat introduced the current status of its participation in the OCUDU project for open source-based telecom infrastructure. The speaker was Red Hat.
The U.S. government officially announced the open source-based open RAN project OCUDU in December last year. Open RAN is an open-interface, standards-based interoperability technology for base station equipment from different manufacturers.
Red Hat is participating in OCUDU as a general member. Red Hat also holds a seat on the OCUDU board.
The technical origin of the OCUDU project is SRS's srsRAN. srsRAN is an open source 4G and 5G software stack that can be used to implement and test software-based base stations.
Red Hat said it has completed the deployment of SRS's initial software stack. Through the initial software stack deployment, Red Hat established a foundation for stable operation of telecom workloads in container environments.
Red Hat said it is also providing its open hybrid cloud platform OpenShift to the telecom industry.
Red Hat said OpenShift comprehensively supports a range of infrastructure, including bare metal and hybrid environments. Photo from the regular meeting of the AINA Industry Convergence Committee on the 29th. The photo was taken by reporter Seong Won-young.
Source: IT DAILY · Seong Won-young
Original: https://www.itdaily.kr/news/articleView.html?idxno=242560
References
This article was produced with the help of an automated content generation algorithm.
Source: IT DAILY
View originalThis article was summarized and organized by BizCrush based on the original article from IT DAILY. For exact quotations and full details, please refer to the original article.