SK Telecom Gave AI Agents Employee IDs. Who Owns Their Data?
An agent with a supervisor still reads data nobody answers for
Most organizations treat AI agent governance as an identity problem. Know which agents exist, give each one credentials, and decide who answers for it.
The larger problem is what those agents read. On October 26, SK Telecom published its first AI Governance Report, which gives each approved in-house agent an employee ID and tracks its tasks, responsible person, operating status and performance, Financial News reported. More autonomous or higher-impact agents must be overseen by a team leader or executive, and the data each agent can access is set by the nature of its work.
That is a serious step, and most companies are behind it. But it puts a recurring question on the table: an agent can now have a named supervisor, while the customer table it reads every hour still has no named owner, no agreed definition and no record of when it was last checked.
And it gets worse as agents multiply. Every new agent needs a data access scope, and every scope assumes someone already knows where sensitive data lives, which copy is current and who can approve access to it.
Agents with employee IDs will soon need to find, verify and trust agents outside their own company, not just the systems inside it. That is what we are building at OpenMap: The Agent Network, the connective tissue for the world's AI agents, starting with every person and every local business. We are raising a £5M seed. Meet us at openmap.ai/investors.
Why Does an Agent With an ID Still Read the Wrong Data?
Because identity tells you who the agent is. It does not tell you what the data is.
Picture a customer service agent at a large retailer. It has an ID, a supervisor and a clear scope: customer records, read only. But the warehouse holds four customer tables. One is the master, one is a marketing copy that drifted months ago, and one is a test extract that still contains real personal data.
Nobody ever recorded which is which.
The agent stays fully within its scope.
It is also reading whichever table it found first.
This is not an agent identity problem. It is a data ownership problem.
The Shortcut Most Teams Take Is the Employee's Permissions
When nobody can describe a scope at the data level, teams borrow one that already exists. In Delinea's fall 2026 identity security report, half of IT leaders said they use the employee's existing permissions as the boundary for agent access, Campus Technology reported.
The result is predictable. 87% said they had experienced or suspected an incident in the past year where an AI tool or agent accessed sensitive information beyond what its task required. Only 19% could detect that kind of out-of-scope access in real time.
An employee's permissions were built for a person who opens a few files a day. An agent can read all of them before lunch.
Before an agent's data access scope means anything, someone has to:
Find every dataset the agent's credentials can actually reach
Mark which copy of each record is the trusted one
Flag which of those datasets hold personal or confidential data
Assign a named owner who approves access to each one
Map which reports, workflows and other agents depend on what it writes
None of these tasks are particularly difficult. The problem is that they are usually done once, for an audit, while the list of agents now grows every month.
How to Give Every Agent's Data an Owner in 30 Days?
You do not need to redesign your data platform or freeze your agent rollout to do this.
Within the first month, focus on the datasets your highest-impact agents already reach: the ones that handle customers, sensitive data or live transactions.
Start by discovering what each agent credential can reach and resolving duplicate and conflicting records, so there is one trusted version of each entity. Then assign an owner and a policy to every dataset on that list.
Next, expose this context to the same approval process that grants agent access, so a scope names trusted, owned datasets instead of borrowing a person's folder permissions.
DataManagement.AI accelerates this process by automatically discovering metadata, mapping lineage across the modern data stack, identifying downstream dependencies, and creating a centralized knowledge layer that stays continuously updated as your environment evolves. It resolves duplicate and conflicting records, and ownership and policy travel with every entity, so an agent's scope points at data someone actually answers for.
Instead of reviewing agent access line by line at audit time, your teams can grant scopes with confidence, because every dataset inside them already has an owner.
Agent Accountability Is Actually a Data Ownership Problem
Giving agents employee IDs is the right instinct. It makes every agent someone's responsibility.
But an agent is only as accountable as the data it acts on. A supervisor can answer for what an agent did. Without data owners, nobody can answer for what it read.
Organizations that scale agents successfully do not just give every agent an owner. They give every dataset those agents touch an owner too.
Warm regards,
Shen and Team