Factmata builds an NLP and machine‑learning platform that classifies misleading, opinionated, or harmful content using eight proprietary detection algorithms (clickbait, hate speech, racism, etc.). The team is working with subject‑matter experts to create training data and says expert annotation yields inter‑annotator agreement above 68%. Its U.S. English news classifiers evaluate at over 90% accuracy on Factmata’s test sets, and the company is expanding to short‑form content and additional languages (starting with French). Factmata has run case studies with 14 major ad exchanges and found up to 4.92% of sampled pages contained high levels of hate speech or hyperpartisan language. The company is taking over and integrating eyeo’s Trusted News Chrome extension (currently “tens of thousands” of users) and plans browser and platform expansions plus more interactive, user‑driven rating features. Financially, eyeo has made a cash and asset investment (terms undisclosed); Factmata previously raised $1 million and has not yet taken VC funding. Factmata is a London-based startup building AI tools to detect biased, incorrect and spoof news content across the media ecosystem. The company is piloting services rather than shipping a finished product and is developing three product strands: adtech tools for programmatic advertising, detection for aggregation platforms, and a consumer-facing extension. Its adtech work aims to detect anomalies and spoof websites and is being tested with partners including AppNexus, Trustmetrics, Sovrn and Bucksense. For aggregation platforms Factmata intends to flag biased or incorrect content to readers rather than remove it. The team is led by CEO and founder Dhruv Ghulati and is pursuing a community-driven AI approach to gather training signals. Factmata says it is taking a long-term (five- to ten-year) perspective on building a new kind of news-safety platform.

Total raised

$1M

Funding rounds

2

Latest round

Seed

Latest activity

Feb 2018