The HWP format used by Hancom Office — which President Lee Jae-myung singled out in his first post-inauguration briefing as ill-suited for AI applications — is effectively being phased out.
On April 24, the National AI Strategy Commission announced that, together with the Ministry of the Interior and Safety and the Ministry of Culture, Sports and Tourism, it will restrict HWP file attachments on core public-document channels such as the Onnara System, citing poor AI recognition of the format.
OpenAI recently added HWP-file support to ChatGPT, slightly improving usability. Still, unlike the open HWPX standard, HWP is a closed format that hampers AI systems’ ability to parse and learn from its internal structure.
Accordingly, the Onnara System — the government’s primary drafting and distribution channel — will extend its mandatory switch to open file formats to local governments beginning on the 18th of next month. The internal communication tool Onmail will complete its transition to open formats by October.
The Integrated Public Servant Mail service, which officials use to communicate with the public, will begin enforcing attachment restrictions in October after a five-month grace period.
The Ministry of the Interior and Safety said it will work with Hancom to ensure that existing HWP documents are converted to HWPX when they are edited or re-saved.
Im Mun-young, vice chair of the commission, said the measure marks the start of small but consequential, fast-moving changes aimed at driving public-sector data innovation in the AI era through interagency cooperation.
Last December, at the 2026 briefing held at the Sejong Convention Center, President Lee received a report from Ahn Hyung-joon, director of the National Data Office, and raised concerns about the Hangul Word Processor’s compatibility with public data, urging concrete countermeasures.
“As data grows in importance, it becomes the core of an AI-driven society,” the president said. “What matters is what high-quality data we produce and how we use it.”
He added, “Government records are among our highest-quality data assets, yet most are written in HWP and employ formatting techniques that machines cannot read. How do we resolve that?”
