Category: Identity | Reading time: 3 min
SK Telecom, South Korea's largest mobile carrier, published its first AI governance report on October 26. Its core idea is simple: an AI agent working inside the company is managed like a member of staff, with an ID, an accountable human, defined permissions and a record of what it does.
How the system works
According to the 40-page report, every in-house agent gets a basic ID. Agents that pass a separate review receive an employee number. Each agent is assigned an owner based on its impact and autonomy, with a team leader or executive responsible for the riskier ones. Permissions and data access are set by the work the agent does, and human approval requirements get stricter as autonomy rises.
The company's Agent OS records the models, tools, execution logs, token usage, cost and changes for each agent, and can halt an agent's execution. SK Telecom has also extended its ISO/IEC 42001 AI management certification scope to cover internal agent management.
The roadmap
The report lays out a phased plan: foundations in 2026, differentiated permissions in 2027 and full lifecycle management in 2028. SK Telecom said it is also considering performance KPIs for agents. It did not disclose how many agents it runs.
Why it matters
Most enterprises still deploy agents under shared service accounts or a developer's credentials, which makes it hard to answer who is responsible when an agent acts. SK Telecom's report is one of the first large-company templates that answers that question in writing: every agent has an identity, an owner, scoped permissions, a log and a stop button. Expect regulated buyers and auditors to start asking vendors for the same.
Source: Financial News, Seoul Economic Daily, TSN Media