Mistral AI Partners with AFP to Enhance Chatbot's Factual Accuracy

GenAI deals with news giants increase as Mistral partners with AFP

The announced comes a day after OpenAI and Google unveiled their own new media collaborations Yet another deal has been signed between a publisher and a GenAI leader. Paris startup Mistral and news outlet Agence France-Presse (AFP) announced today that they are combining their services to improve AI responses. The deal provides Mistral's chatbot -- imaginatively named Le Chat -- with access to all of AFP's text stories. According to Mistral, the integration will bring "enhanced factuality" to the AI assistant. "Partnering with a globally trusted news agency like AFP allows Le Chat to offer reliable, factual, and up-to-date responses, verified by professional journalists," said Arthur Mensch, the startup's CEO and co-founder. Mistral also highlighted the linguistic capabilities. AFP's daily production of 2,300 text stories spans six languages -- French, English, Spanish, Portuguese, German, and Arabic -- all of which will be available to LeChat. "Through this partnership, we are providing our clients with a unique multicultural and multilingual alternative," Mensch said. For AFP, the deal adds an extra income source and a new outlet for the agency's 1,700 journalists. "Through this partnership, AFP is further diversifying its revenue sources, reaching a clientele beyond the media sector and exploring new uses for its content in the daily operations of businesses," said Fabrice Fries, the company's CEO and chairman. Deals and disputes between GenAI firms and news outlets The content deal expands a growing range of agreements between publishers and GenAI companies. Just a day earlier, two other eye-catching collaborations were announced. One brings news from the Associated Press (AP) to Google's Geminis chatbot. The other merges Axios journalism with OpenAI products. As part of the deal, the ChatGPT-maker will fund the media brand's expansion to four new US cities. Axios joins a lengthy list of publishers collaborating with OpenAI. The GenAI giant has now partnered with nearly 20 media organisations, including the Financial Times, Le Monde, and AP. Yet not every news outlet has been a willing collaborator. A group of them -- led by The New York Times -- took OpenAI to court this week over alleged copyright infringements. They argue that OpenAI used their content to build systems without consent or payment. OpenAI contends that the "fair use" law protects the practice. By signing agreements with publishers, GenAI firms could avoid such legal disputes. For Mistral, the partnership with AP is the first content deal of this kind. The integration is slated to roll out to all Le Chat users in the coming weeks. Neither Mistral nor AFP have revealed the value of the multi-year contract.

Mistral signs deal with AFP to offer up-to-date answers in Le Chat | TechCrunch

This is the first deal of this kind for the Paris-based artificial intelligence company. And it indicates that Mistral doesn't want to be considered as "just" a foundation model maker. It also wants to build appealing products, starting with Le Chat. From what I've heard, the company is also working on dedicated apps to access Le Chat and better compete with ChatGPT or Claude. Going forward, Le Chat will be able to tap into AFP's daily production of stories. And given that AFP is one of the biggest news agencies in the world, it represents a significant volume of text -- around 2,300 stories per day in six languages (Arabic, English, French, German, Portuguese, and Spanish). Le Chat will be able to query AFP's entire archive since 1983. However, photos and videos aren't part of this multi-year agreement. As a reminder, Mistral focuses on large language models and doesn't offer image-generation models. OpenAI has been leading the charge when it comes to content deals. The maker of ChatGPT has inked partnerships with AP, Axel Springer, Condé Nast, El País, Financial Times, Le Monde, and others. It's going to be interesting to see whether Mistral has more content partnerships in the works. "We believe improving the accuracy of [Le Chat's] responses is a key step in the deployment of our technology, particularly for businesses," Mistral co-founder and CEO Arthur Mensch said in a statement. "Through this partnership, we are providing our clients with a unique multicultural and multilingual alternative." Today's partnership is also a first for AFP. And it couldn't come at a better time, as Meta ended its third-party fact-checking program just last week. AFP was one of the key partners in Meta's fact-checking system. "Through this partnership, AFP is further diversifying its revenue sources," AFP chairman and CEO Fabrice Fries said in a statement. While the AI industry is looking to improve its products with these arrangements, there are two side effects that could be considered as added benefits. First, AI companies can position themselves as (financial) allies to news organizations. Second, these partnerships protect them from potential copyright infringement claims.

Mistral signs AFP deal for fact-based chatbot in riposte to 'free speech' rivals

French artificial intelligence start-up Mistral has struck a multimillion-euro deal with Agence France-Presse to incorporate thousands of the newswire's articles into its chatbot, pitching the tie-up as a European bulwark against attacks on fact-checking from its Silicon Valley rivals. The partnership between AFP, one of the world's oldest news agencies, and Mistral is the first of its kind for the two Paris-based companies, when many media groups are deciding whether to strike licensing agreements with AI companies or take legal action over alleged copyright infringement. The deal, announced on Thursday, will feed more than 2,000 AFP news articles in six languages every day into Mistral's chatbot, Le Chat, allowing users to answer questions and help draft documents. "It's important to have such agreements to have well-grounded information on validated content," Arthur Mensch, Mistral's co-founder and chief executive, told the Financial Times. The companies presented the deal as a means of ensuring Mistral's chatbot is grounded in verifiable information. It comes as Meta and Elon Musk's X have pulled back on content moderation and declared the primacy of "free speech", in the run-up to incoming US president Donald Trump's inauguration. "What it tells us is that Europe must unite to defend its thriving technological sector," Mensch said about recent moves by Silicon Valley rivals. "'Free speech' is being weaponised against Europe to a great extent and there is this offensive by Big Tech on European regulation," AFP chief executive Fabrice Fries told the FT. "Precisely this kind of deal, in the current context, shows that an AI player has bet on independent, fact-based professional journalism." On Wednesday, Google announced a similar deal with Associated Press, a longstanding partner on its search engine, to show the newswire's feed in its Gemini AI app. Mistral raised €600mn in new funding at a €6bn valuation in June last year, making it Europe's most prominent AI company and the continent's only start-up making large language models that rival the likes of OpenAI, Anthropic and Elon Musk's xAI. Mensch said Mistral offered a partnership model that was "more open" and "shares the value more evenly" than its US competitors. Fries said AFP had discussed licensing deals with several AI companies in recent months, "but it's only with Mistral that we have had the feeling that it was a genuine partnership, not just a sale agreement". Commercial terms of Mistral and AFP's deal, which runs over multiple years, were not disclosed. But unlike similar agreements struck between US-based OpenAI and other media groups, Fries said the deal was "not a one-off settlement" for data on which large language models are trained. OpenAI has struck content deals with media groups including News Corp, Axel Springer and the Financial Times. On Wednesday, the San Francisco-based group led by Sam Altman said it would fund four new local US newsrooms for online publisher Axios, with the output feeding into ChatGPT. Fries said dealing with AI companies was "still an open battle" and that he was closely tracking the US legal case between OpenAI and the New York Times over copyright infringement claims, which is set to offer a new precedent on the value of the work by publishers to AI model groups. For AFP, the deal with Mistral also represents an opportunity to make up revenue that will be lost as its fact-checking contract with Meta winds down. The US social media group said last week that it planned to shift to community-based fact checking in the US. AFP has 150 journalists working for Meta on fact checking, according to Fries. AFP made about €20mn in 2024 from tech platforms, including fact checking for the likes of Meta and content licensing deals with platforms including Google, accounting for about 10 per cent of its commercial revenues last year. "Now clearly this pocket of revenue which has helped us grow and show profits in the past seven years is at risk," Fries said. "We clearly need to find new tech players as a source of revenue and AI actors can be a replacement for the platforms."