The government decided to reduce existing HWP attachments in public document distribution channels such as the On-Nara system starting from May 2026. Instead, it plans to increase the use of HWPX, an open document format. This policy is being pushed together by the Ministry of the Interior and Safety, the Ministry of Culture, Sports and Tourism, and the National AI Strategy Committee. The government explained that HWP has a structure that is not good for AI to read and analyze. On the other hand, it sees HWPX as better for AI use because information inside the document can be taken out more easily. So it decided to guide even existing HWP documents to be changed to HWPX when they are rewritten or saved after editing. Recently, ChatGPT started supporting HWP reading, but the government judged that changing to an open format is more important in the long term. This measure is part of a broader move to change public documents from files that people read into data that machines can also read.
원문 보기
Why the government is now even fixing file formats
At first glance, this news looks like it is just about changing one file extension. But if you look a little closer, it is more like a signal that the government wants to move from document administration to data administration. It means going beyond the time when people just open and read documents, and making it possible for AI to read, search, summarize, and connect document content in a structured way.
An important term here is machine readability. Simply put, it means how well a computer can understand not just the outer appearance of a document, but also where the title is, where the main text is, and where tables and lists are. If you understand this, you can see why the government is not trying to ban HWP itself, but instead reduce the share of HWP in public systems and move to open formats like HWPX.
Another thing to look at is the timing. People have kept talking about the digital platform government, the use of public data, and AI administration, but if documents are still in a format that is hard for machines to read, they cannot move to the next step. So you can see this measure as changing document writing habits, and at the same time rebuilding the basic foundation for connection and automation between ministries.
The key is not that 'Hangul files are bad,' but that they want to create a public document structure that is easy for AI and systems to read.
You can see this measure not as the stage of making more electronic documents, but as the stage of turning electronic documents into usable data.
HWP and HWPX may look similar outside, but the inside structure is different
| Criteria | HWP | HWPX |
|---|---|---|
| Storage method | Traditional binary based(a parser is needed rather than being directly easy for machines to read) | XML-based package structure(easy to extract document elements by rules) |
| Whether the structure is open | Relatively closed | Open structure based on the national standard KS X 6101 |
| How easy it is for AI to read | There are many extra interpretation steps to extract text and structure | It is relatively easy to find structure clues like paragraphs, tables, and lists in a stable way |
| Conversion and linking | When connecting with other systems or tools, the burden of special processing is high | It is relatively easy to connect with general XML and compression tools |
| Administrative meaning | It is strongly compatible with existing practices | Good for AI use, interoperability, and long-term standardization |
This is how AI reads documents
"Opening a file" and "understanding a document" are completely different things. If you look at this process, you can quickly understand why HWPX is better.
Step 1: Open the file
The first thing needed is to technically open the file. HWP can also be opened with special tools, but HWPX is a combination of a ZIP package and XML, so it is easier to access with general tools.
Step 2: Extract the text
AI does not look at the screen. It has to pull out the letters from inside the file. Because HWP is a binary format, the burden of separate interpretation is big at this step, and in HWPX the text is easier to appear in a structured form.
Step 3: Understand the structure
This is the really important part. Search and summary quality get better when it can tell apart titles, body text, tables, lists, and footnotes. HWPX is much better for catching these clues because of XML tags.
Step 4: Reuse it for work
Things like automatic document classification, draft answers for civil complaints, extraction of record Meta data, and reuse between ministries happen at this step. So changing to HWPX is not just making files look neat. You can think of it as changing the materials for administrative automation.
Why did Korean public documents become centered on HWP
If you know this, instead of asking "Why change it only now?" you can better see "Why was it kept for so long?"
Step 1: In the 1990s, Araeahangul matched the Korean language environment
In the early PC environment, Korean input and document layout quality were very important. At that time, Araeahangul spread quickly and began to settle as the basic document tool for schools, companies, and public institutions.
Step 2: E-government put HWP practices into the system
The problem did not grow only because it was popular. Electronic approval, electronic documents, internal forms, and document exchange between institutions were designed around HWP, so the format choice became built into the work process itself.
Step 3: Standardization discussions started, but inertia was stronger
In 2011, OWPML was established as the national standard KS X 6101, and discussion based on an open structure entered the institutional system. Even so, in real workplaces the HWP-centered ecosystem did not change easily because of existing forms, budgets, training, and compatibility issues.
Step 4: As the AI era arrived, the limits started to stand out more than old habits
In the past, there were many documents that only people needed to read, but now systems and AI also need to read documents. So old practices that lasted a long time have now started to look like barriers to efficiency and connection.
It was not only Korea's issue: open document changes by country
| Case | What they did | Lesson left behind |
|---|---|---|
| Massachusetts, United States | In the mid-2000s, it pushed a policy centered on open formats like ODF and PDF | It is not enough to only set principles. You also need to design compatibility with existing work and a transition plan together. |
| UK government | In 2014, it adopted ODF as the standard for sharing government documents | It clearly showed the principle that public documents should be openable by anyone even without a specific company's program |
| EU | It connected interoperability and avoiding vendor lock-in with public procurement principles | It is important that the document format issue is seen as part of a broader digital government strategy |
| Korea | It kept the HWP ecosystem, then expanded discussion on OWPML standardization and switching to HWPX | While respecting local practices and compatibility, the need for an open transition is growing because of AI, long-term preservation, and connectivity |
Reality of civil servant document writing: most formats are still hard for AI to read
What will change first in the field from now on
Just because a policy was announced does not mean all old files change overnight. The real transition usually goes like this.
Step 1: HWPX will increase first in newly shared official documents
The first change people will notice is in new documents like notices, application forms, and attached templates. When downloading, there will likely be more cases where the file extension changes to HWPX first.
Step 2: Existing HWP files are also gradually converted during editing and saving
The government plans to encourage even existing HWP files to be saved as HWPX when they are rewritten or edited. In other words, rather than converting old documents all at once, it is closer to changing little by little, starting with active documents.
Step 3: From October 2026, mandatory basic format use begins, and old version users may get confused
According to news reports, the government has suggested a plan to make HWPX mandatory as the basic document format for public institutions starting in October 2026. This can cause confusion in actual work. In particular, users of the 2010~2018 versions need updates, and even older Hangul programs may not open HWPX right away, so workarounds like installing a viewer may become necessary.
Step 4: After automatic conversion, checking and keeping both versions becomes important
For documents with many tables, stamps, formulas, and complex formatting, it is hard to rely only on automatic conversion. So for a certain period, there is a strong chance of transition costs that include keeping the original, checking the converted version, and doing security inspection together.
So how should we read this news?
If you read this news as just 'the government started to dislike Hangul files,' it is easy to miss the main point. A more accurate reading is that this is an attempt to change public documents from results that only people read into input data that AI and systems can also read.
From this view, the issues to watch in the future also become clear. First, how far the HWPX conversion really goes in the field. Second, how well compatibility and formatting quality problems are managed during the conversion process. Third, whether changing the document format really leads to better administrative efficiency, like search, summary, and automation of civil complaint processing.
If you understand up to here, the next time similar policy news comes out, you will feel much less lost. Even if the story is about file extensions, now you can also look at the real points behind it, like standards, interoperability (the ability of different systems to connect without problems), and machine readability.
See whether the new document extensions distributed by public institutions really change mainly to HWPX.
The success of the transition will likely be decided more by compatibility management in the field and results from using AI than by declarations.
We will tell you how to live in Korea
Please give lots of love to gltr life