Be aware of harmful claims and narratives circulating online
Industries and organisations need to know what is circulating online. Messaging apps, websites and research documents produce text at a scale humans cannot review and LLMs cannot effectively handle.
Our models process millions of sentences across 114 languages, surfacing disputed narratives and controversial claims judged against your trusted sources.
Our models process millions of sentences across 114 languages, surfacing disputed narratives and controversial claims judged against your trusted sources.
text stream resolving into Noteworthy claims
Collected
114 languages
Verified
412
63
190
.png)
.png)
Defence, intelligence, and media intelligence engagements — proof line to be pinned to a named engagement.
01
Direct your resources to the most important claims and narratives
If you are gathering and analysing large amounts of text, the first challenge can be figuring where to focus your attention.
Factiverse finds the controversial claims inside millions of sentences in minutes. This enables analysts and researchers to spend more time on judgement on what is relevant for their reporting.
Factiverse finds the controversial claims inside millions of sentences in minutes. This enables analysts and researchers to spend more time on judgement on what is relevant for their reporting.
02
Tailor your verification pipeline with sources you trust
After each claim identified, it comes back with evidence that supports or disputes it from sources that your teams trust.
You see at a glance what is corroborated, what is contested, and what rests on a single unreliable source. This informs teams on how reliable call each claims before it reaches them.
You see at a glance what is corroborated, what is contested, and what rests on a single unreliable source. This informs teams on how reliable call each claims before it reaches them.
03
LLMs cannot replicate this process
Running open-source volume through a large language model is too slow and too costly to sustain. Coupled with hallucinations this makes it unfeasible to effectively tackle textual information.
The Factiverse claim detection and verification models are built for exactly this. Verifying continuously across 114 languages at a cost that makes it viable.
The Factiverse claim detection and verification models are built for exactly this. Verifying continuously across 114 languages at a cost that makes it viable.
Case study
We gave a NATO member state's analysts daily claim monitoring across Telegram in every language they needed
This case study discusses how a NATO defence force turned billions daily Telegram messages into a searchable claim database to assit their daily reporting.
100+
Telegram channels monitored and analysed daily.
114
Languages monitored across sources for teams who do not native speaker analysts.
24h
From message posted to claim surfaced in the daily roundup.
Turn open-source volume into verified intelligence
Book a consultation and Factiverse will scope verification around your sources and your tasking.
Frequently asked
Do you collect the text, or do we?
You collect as you do now, with whatever tools you already use. Factiverse takes the text you have gathered and verifies it. Collection stays yours; assessment is where we add value.
What kinds of text can it handle?
News and wire copy, social media and forum posts, blogs, reports, and research papers. If it is text, Factiverse can find the checkable claims inside it.
How do you verify a claim?
Each checkable claim is tested against the FactiSearch database and marked by the sources that support or dispute it, so you see what is corroborated and what rests on a weak or single source.
Why not just use an LLM for this?
LLMs work at small volume, but not at the throughput open-source work demands, where they become too slow and too costly. The Factiverse claim detection model is built for that scale.
Can it work in more than one language?
Yes. Verification runs across 114 languages, so foreign-language sources are assessed alongside English rather than left in a backlog.
More on verification at scale
.png)
By the way
Subscribe to our newsletter to receive important developments about everything AI safety related
Discover how Factiverse solutions can help The Research Video and AUDIO for your teams