Does Flitto: Multilingual Data for AI & Real- have a moat?

Flitto says it has 14M users in 173 countries. Since 2018 its Flitto Arcade app has turned translation, transcription and voice recording into small paid tasks, with 3 to 7 layers of review that rank contributors by past quality.

To cross it: Pay to recruit and vet native speakers in long-tail languages through gig platforms, and build up quality scores on each contributor over many months.

Flitto has passed the test batches and legal-clearance checks of US tech giants. In 2025 it disclosed contracts with global IT clients that later grew to a combined 31.6B KRW, and revenue rose 77% to 36B KRW with its first annual operating profit (6.2B KRW).

To cross it: Win small pilot datasets from labs and deliver accuracy and licensing audit trails good enough to get onto their approved-vendor lists.

Flitto has 13 years of crowd-made parallel corpora and speech in 90+ languages, cleared for commercial training. It has also taken part in the National Institute of Korean Language corpus project five years running.

To cross it: Build fresh, licence-clean datasets through paid collection or synthetic generation in the language pairs buyers want right now.

Its interpretation product runs at Incheon Airport, Seoul Station, Shinhan and Hana banks, CU stores and a Hanwha Ocean shipyard. It is also tuned to each site's terms, and a 3.87B KRW contract with a semiconductor firm covers similar work.

To cross it: Offer the same thing cheaper with custom glossaries, and win the next public tender or enterprise renewal.