The US trade association News/Media Alliance (NMA) has partnered with GenAI platform Bria to let members opt into a licensing agreement that would let their content be used to train and surface in AI systems.
The partnership will form the bedrock of a new AI product from Bria, giving the company access to content from some of NMA’s 2200 news, magazines, and online media outlets.
Bria says it has created an attribution technology that assigns credit to a licensed piece of copyrighted content, meaning that when content is used in its tools, it can appropriately compensate publishers.
The move would give participating publishers a recurring revenue stream when their content is used in Bria’s products.
Publishers have been immensely impacted by the rise of GenAI models like ChatGPT and Claude. Much of their traffic has been taken by AI models, often with their own content being surfaced in these models without permission or compensation.
This style of agreement serves as a possible antidote to these impacts, which have shaken up the editorial industry and its business model.
“The frequent use of publisher content by AI companies without compensation has been devastating for our industry, and we are pleased to offer a reasonable, responsible alternative,” explained Danielle Coffey, President and CEO of the News/Media Alliance.
Content for RAG
AI companies are always in search of high-quality content to train their models, because the better the data used to train the models are, the better the AI-generated output.
This type of agreement ensures AI companies, like Bria, can provide its customers with quality AI-generated outputs with more accuracy and fewer chances of surfacing erroneous information.
In part, this is because AI companies use a framework called retrieval augmented generation (RAG), which involves the AI model fetching trusted external data – such as the content from NMA’s publications – to surface outputs with fewer errors or hallucinations.
In a statement, NMA said it will join a growing movement of publishers selling their content to AI companies for frameworks like RAG.
“By giving AI companies the opportunity to reach multiple publishers at once, we are making it easy for them to reduce transaction costs and responsibly source their content,” continued Coffey.
This is particularly important because of the prevalence of hallucinations in AI models – where the AI generates fabricated or incorrect information that often sounds believable.
