US Midterm elections, Anthropic integrates Claude chatbot against fake news

Information certified against the limits of AI

The breakthrough responds to a structural problem of language models. Typically, when a user asks a chatbot where to vote or which documents to present, the AI ​​draws on general knowledge that risks being incomplete. Thanks to this partnership, Claude will directly query the Democracy Works database, which collects and certifies official data from state and local election authorities.

The weight of politics in user chats

The operation is based on a strong growth trend analyzed directly by the company through a study conducted on 250 thousand interactions in the United States. The data shows that total conversations on Claude.ai regarding political topics, public policies or social issues stand at 5.75%. 0.51% of conversations focus specifically on elections and voting procedures, a figure Anthropic expects to rise sharply as Election Day approaches. On the contrary, direct requests on who or how to vote remain a completely marginal share, equal to just 0.01% of the total.