Gov AI detects check-worthy claims across social, messaging, press and broadcast lanes, groups them into narratives and claim families, and traces where they started and how they spread. Analysts decide whether to respond or hold, a person approves every response, and it is published openly from the official account. The platform never creates fake accounts, runs covert bots or amplifies content inauthentically.
Claims are raised from collected posts, messages, articles and broadcast transcripts, grouped into families of the same claim in different wording, and triaged to respond or hold with the reasoning recorded.
Narratives are tracked across lanes with their growth over time, and a change radar shows how wording shifted after an official reply. Origin tracing orders near-identical posts by time to show the earliest collected appearance, the spread, platform hops and wording drift, with the limits of the collection stated.
Accounts that reuse wording, post in lockstep or amplify together are grouped and scored, and an analyst rules on each cluster. Writing-style comparison across accounts is shown as an automated inference, never as proof of identity.
Claims are checked against retained sources. A citation dependency map shows which reports and replies point back to the same material, and the recycled media inspector shows when the same image or video reappears with a new caption or a new date.
Before release, the stress lab binds each sentence of a draft to the source passage behind it and flags figures, qualifiers, negations and quotations that could lose meaning when excerpted. After publication, an observation lab compares collection in equal windows before and after, without claiming causation.
Responses are drafted from the evidence, routed through an approval policy and published from the organization’s own attributed accounts. An automated run can collect, detect and draft at volume, but nothing reaches the public without a person’s approval.
No. The platform never runs covert bots, fake accounts or inauthentic amplification. Responses are published openly from an organization’s own official accounts, and only after a person has approved them.
Authorized people in your organization, under an approval policy that can require two or three different approvers. Role-based access controls who can draft and who can approve, and each step from draft to publication is recorded in the tamper-evident audit trail.
Not on its own. The earliest post the platform collected is not necessarily the original, and coordination is an inference about real accounts. Findings are presented with their evidence, the analyst’s confidence and the limits of the collection, so attribution is argued rather than assumed.
Yes. Claim and narrative detection run on the same multilingual collection as social media intelligence, which covers 11+ languages including isiZulu, isiXhosa, Sesotho, Shona and Portuguese, and on radio and television transcripts.
See counter-disinformation working on your own sources, in a secure environment with clearance verification.