AFP partners with Mistral AI to enhance news delivery through chatbot technology

Global news agency Agence France-Presse (AFP) and French artificial intelligence start-up Mistral AI have announced a partnership that will see Mistral’s chatbot, Le Chat, utilise AFP’s journalistic reports to respond to user inquiries. The announcement was made on Thursday, although the specific financial details and duration of the contract were not disclosed.

This marks the first collaboration of its kind for AFP, positioning Mistral AI as a notable competitor to major American AI companies like OpenAI, creator of ChatGPT. Although partnerships between news organisations and AI developers are still uncommon on a global scale, there has been an increase in such agreements since the previous year. OpenAI has formed a number of relationships with various media outlets, including the Financial Times in the UK, Le Monde in France, and Germany’s Axel Springer.

Fabrice Fries, the CEO of AFP, spoke to AFP journalists, highlighting the significance of the partnership, stating, “This is the first deal between two players with global ambitions, indeed a global footprint as far as AFP is concerned, but with well-anchored European roots.” Fries also noted that this initiative will generate a new revenue stream for the agency.

From Mistral AI’s perspective, the collaboration is a strategic move to enhance the value of its offerings. Founder Arthur Mensch conveyed that AFP provides “a verified, journalistic source that we think is very important,” which adds credibility to their chatbot.

Starting Thursday, users will be able to access AFP’s articles in six languages—French, English, Spanish, Arabic, German, and Portuguese—through Le Chat. The chatbot functions similarly to ChatGPT, where users can input questions and receive responses in a matter of seconds. Le Chat will answer inquiries about current events using AFP’s text news content, which typically gets distributed to clients across various sectors, including media, government, and business. During an initial test phase, access to this feature will be limited to a select group of Mistral users.

The chatbot can reference an extensive archive of about 38 million articles dating back to 1983, with AFP releasing approximately 2,300 new articles daily. Fries posited that access to this information could be particularly beneficial for “professionals or managers in large businesses” who may require assistance in drafting memos or documents concerning current affairs.

Complementarily, the broader public is increasingly engaging with generative AI tools for various purposes, often seeking information on everyday topics. Mensch highlighted the distinction between the types of data sourced by the chatbot, stating, “Where users’ questions require verified information, AFP will provide the inputs. Concerning shopping or the weather, it will come more from the web.”

The timing of the AFP-Mistral partnership coincided with recent announcements from Meta, the parent company of Facebook and Instagram, which indicated a cessation of its fact-checking programme in the United States. Despite speculation, Fries clarified that AFP’s discussions with Mistral began nearly a year ago and were not directly influenced by Meta’s decision.

As traditional media grapples with ongoing challenges, AFP continues to explore diverse partnerships with digital platforms. The agency recorded its fifth consecutive annual profit in 2023, generating €1.1 million, supplemented by €113.3 million from the French government for public-interest objectives out of a total revenue of €320.1 million.

In contrast to other media-AI collaborations, AFP’s text articles will not be utilised for developing Mistral’s language models. Instead, the agency’s content will operate as “a module that connects to our system and can be disconnected” following the contract’s termination, according to Mensch. This arrangement is framed as a means of creating “recurring revenue” for AFP, as Fries explained, “This isn’t a one-and-done payment, as is often the case in deals for training models, but development of recurring revenue.”

Source: Noah Wire Services