https://ifpim.org/resources/new-partnership-launched-to-ensure-linguisti...
Paris, 10 February: The International Fund for Public Interest Media (IFPIM) today announced the creation of the Global Media Trust, a new initiative and platform to enable news content creators - particularly those producing content in the Global Majority - to be paid for content essential to training AI Large Language Models (LLMs) and to foster linguistic diversity in the development of AI technology.
The announcement comes during the Artificial Intelligence (AI) Action Summit, where Heads of Government, the Private Sector and NGOs are gathered this week to work towards global, inclusive and equitable development of AI technologies to serve the public interest.
“The Global Media Trust initiative is designed to ensure that users of AI models receive accurate, verified and balanced information, regardless of their preferred language. To do this, we need to solve two current gaps in how LLMs and generative AI tools are currently being developed,” said Nishant Lalwani, CEO of the International Fund, announcing the new initiative. “First, this Trust will address the lack of linguistic diversity of AI tools, which disadvantages the non-English speaking majority of the global population. Second, it will create a mechanism for Big Tech AI developers that need journalism content and news media producers to come to agreements that benefit them both. The Global Media Trust offers a win-win solution, not least for audiences that have a right to trustworthy information.”
The International Fund is a global financing mechanism created to support independent media in low- and middle-income countries through direct financial support, support for innovation, and efforts to secure the future of public interest media through developing new business models for the sector.
While more than 7000 languages are spoken worldwide, the vast majority of AI technologies and LLMs are being trained and optimized for use in English, leading to low quality, distorted or bias-reinforcing information generated by AI tools in ‘low-resource’ languages. Studies show that the lack of linguistic and cultural diversity causes LLMs to be prone to delivering more misinformation, hallucinations, and biases when operating in low-resource languages. This gap risks worsening existing inequalities in access to trustworthy information.
Groundbreaking deals struck in 2024 to pay news platforms for their content have thus far mainly been with entities owning English language news platforms: OpenAI’s agreements with the Associated Press, as well as Meta’s deal with Reuters both focus on English, while a deal with Axel Springer include English-language as well as German content. Only a few agreements such as the deal between OpenAI and Le Monde focus on non-English content.
To facilitate equitable remuneration for content, the Trust will broker collective and individual licensing opportunities with AI companies for public interest information creators in the Global South.
The Trust is also designed to foster localized development of AI technology, leveraging open-source and low-cost LLMs to develop generative AI tools optimized for local communities that have been traditionally underserved by global tech platforms. This will allow coalitions of public interest information creators to build and own platforms to best serve users in their local markets.
The GMT has been developed in consultation with Current AI, which is a flagship deliverable of the AI Action Summit in Paris on February 10th and 11th, 2025.
For more on the AI Action Summit please see: https://www.elysee.fr/en/sommet-pour-l-action-sur-l-ia/presentation
