KBS·MBC·SBS vs OpenAI lawsuit: Practical response guide to Korean media AI copyright war

On February 23, 2026, KBS, MBC, and SBS filed a lawsuit against OpenAI regarding unauthorized use of news content. This incident is not just news of a dispute, but a turning point where domestic media companies and AI product teams must redefine “What can be used as learning data” .

1) Define the problem: Who needs to make what decisions now?

Target readers: ​​Media digital/legal organizations, AI service planners, data governance personnel.

Problems to be solved: ​​How to reduce legal risks when using news content in model learning, search augmentation (RAG), and summary functions.

Scope: ​​News content usage policy, licensing agreement, internal control process.

Exclusion range: ​​The purpose of the text is not to provide legal advice for a specific lawsuit, but to present a practical operating frame.

2) Evidence/Comparison: 3 operating strategies

Strategy	Cost	Execution speed	Accuracy/Quality	Legal Risk
Maintain unauthorized use of data	Short Term Low	Fast	Initial High	Very High
License Agreement Centered	Medium~High	Medium	High	Medium~Low
Public/private data conversion	Medium	Medium~Slow	Medium	Low

The core message of this lawsuit is clear. Post-dispute costs may be greater than the initial cost savings. . In particular, contract-based data strategies have become the default for companies targeting B2B/public markets.

3) Step-by-step implementation guide (for practical purposes)

Step 1. Data source inventory completed within 48 hours

All text sources used in the current service (chatbot, summary, recommendation) are investigated and tagged in three stages: “licensed/unknown/unlicensed.”

Step 2. Immediately quarantine high-risk sources

Content with unknown rights is immediately excluded from the new learning pipeline, and the search/exposure cache is also separated.

Step 3. Design contract priority matrix

Check traffic contribution and possibility of dispute together and switch contracts starting from the top 20% of sources.

Step 4. Force source/date notation on product UI

In the summary/answer function, the original source and publication date are required to reduce the risk of misunderstanding.

Step 5. Quarterly audit routine

The legal, development, and data teams jointly prepare a “Learning Data Audit Report” once a quarter and report it to management.

4) 4 common traps and recovery methods

• Plot 1: Misunderstanding that “learning is possible on the open web” → Repair: Document terms of use/copyright policy by source.
• Trip 2: Judging that it is okay because it is a PoC. → Recovery: Apply the same rights verification checklist from the PoC stage.
• Trap 3: Shifting responsibility to model supplier → Recovery: Specify data responsibility sharing clause in contract.
• Trap 4: Reduced trust by not indicating the source → Recovery: Automatically insert the source and date into the answer template.

5) Execution checklist + DoD

• Data source 100% inventory complete
• Apply unknown rights source isolation flag
• Top risk source contract negotiation list confirmed
• Indicate source/date in product response Release
• Quarterly audit template and responsible person designation
• Distribute playbook to respond within 24 hours when infringement issue occurs

Definition of Done: Complete when confirming with the supporting log that “data with unknown rights does not remain in the production learning/response path”.

6) Reference material (source + date)

• AI Times - KBS, MBC, SBS files suit against Open AI for 'unauthorized use of news content' (2026-02-23)
• Electronic News - Report on lawsuit against OpenAI by three terrestrial broadcasters (2026-02-23)
• Maeil Business News - Broadcasting company-OpenAI copyright dispute report (2026-02-23)

7) Author's perspective: “Contractable expansion” is now more advantageous than “rapid expansion”

For domestic AI service providers, the deciding factor in 2026 is rights consistency and auditability rather than model performance itself. Even if costs increase in the short term, organizations that preemptively establish a contract-based data system are likely to have a long-term advantage in regulation, litigation, and enterprise sales.

There is an exception. Some mitigation is possible in in-house-only, non-commercial, closed laboratory environments. However, the moment it is converted to an externally exposed product, it must be immediately upgraded to the same standard.