SK电讯发布首份AI治理报告，将内部AI智能体纳入类似员工的管理与绩效框架

SK Telecom (017670) is formally recognizing artificial intelligence agents as a new form of labor and is reviewing plans to reflect their performance in the key performance indicators, or KPIs, of its organizations. The carrier also said it will tighten human review and approval procedures for AI agents with greater autonomy, and raise its level of management and oversight by examining each agent's token usage and work records.
SK Telecom said on the 26th that it had published its first "AI Governance Report," covering its AI governance principles, accountability framework, specific action plans and results. The report was produced to further strengthen "AI trust management," an approach aimed at improving the safety and reliability of AI technology.
In the report, SK Telecom disclosed the current status of and plans for its in-house AI agents. The company assigns employee ID numbers to AI agents as it does to human staff, and continuously monitors each agent's role, designated supervisor, operating status and performance. "For AI agents that have greater impact or higher autonomy, a supervisor such as a team leader or executive — not a rank-and-file employee — must be designated to oversee them," SK Telecom said. "We also operate an agent operating system to manage AI agents developed and run in various environments under consistent standards."
SK Telecom also plans to recognize AI agents issued employee ID numbers as an officially new workforce and to run a system linking their operating performance to organizational performance. "We are reviewing ways to share agents' performance with our employees," the company said. "Over the longer term, we will consider reflecting it in organizational KPIs." Agents judged less necessary based on evaluation results will be suspended or retired.
"In preparation for an environment in which in-house AI agents carry out work more autonomously, we plan to progressively build out a management framework spanning identification of each agent, designation of a supervisor, access control, regular evaluation and monitoring, and suspension or retirement," said Cha Ho-beom, SK Telecom's chief privacy officer.
The report also emphasized principles for protecting customers' personal information. "As a matter of principle, we apply 'Privacy by Design,' which takes personal information protection into account from the planning and design stages of a service, to service planning and development," SK Telecom said. "To that end, we have established and distributed AI guidelines, which we use as a standard for raising the level of both privacy protection and AI security." When an AI service that handles personal information adopts an external model or library, the adequacy of such components — including licensing and security vulnerabilities — is also subject to review.
"SK Telecom reviews the safety and reliability of AI technology from the service planning stage and manages the process so that identified risks lead to actual remedial measures," Cha said. "Going forward, we will upgrade our inspection and monitoring framework across the entire lifecycle so that we can respond to changes in AI models and data, and to the possibility of new risks emerging, even after a service is launched." Chung Jae-hun, SK Telecom's chief executive, said: "We recognize the responsibility that comes with AI as much as the innovation it will bring. We will make sure customers can trust not only the performance of AI but also the accountability of the company providing it."