ALERT: New York World Congress registration deadline is Friday, click here for agenda and to register now

Yes, news companies need AI middlemen (here are 3)

By Jodie Hopperton

INMA

Los Angeles, California, United States

Connect      

I am in the midst of researching different AI futures for news products and our businesses. This includes visits major tech companies in Los Angeles, Silicon Valley, and Seattle on a mission to get to know LLMs and their relationships, or intentions, with news.

As part of this research, I have gotten to know three companies that are acting as what I would call the “middlemen,” trying to find ways for rights holders to be paid for the content being used. 

The need for middlemen

There is a need for platforms to act as middlemen between content owners and LLMs. One-to-one deals don’t scale, and AI only works at scale. There are three companies that are looking to solve similar problems in slightly different ways: how to get rights holders paid by LLMs.

LLMs do not have the teams, and in some (many?) cases the will, to deal with hundreds and thousands of publishers to strike individual deals. It makes no sense to do that. I think we can all agree some kind of framework needs to be established for what monetisation looks like for content owners. None of us are going back to the early Internet days of giving content away for free — especially as advertising revenues and our direct relationship with customers are under threat. 

In many ways, it makes sense for LLMs to pay for content. Other than ethical reasons, reliable content is much easier for them to deal with and pointing to a source of information reduces liability. 

I’ve spent some time getting to know three companies that are focusing on this very idea, all with very different approaches: Human NativeProRata.ai, and Tollbit. Here is my assessment of how each is tackling the problem, what their models are, and who is behind each.

Tollbit 

TLDR: Uses a line of code on your site to measure what is being scraped by whom so you can manage it, block it, and/or get payment for your content.

The big picture: The team believes you have to measure before you can manage. They are giving the code and measurement tool to news organisations for free to build traction and show the extent of the scraping. Assuming the measurement is correct, which I have no reason to doubt, news organisations also get to see what kind of content is most popular — and therefore more valuable. At the news organisation’s request, they change robots.txt to be “active” code that actually blocks LLMs and directs them to a payment mechanism.

The deal: The code is free, and they plan to take fees from LLMs as they buy content. 

The people: The founders are young and smart. Coming from different sides of a similar issue, they have accurately assessed the problem(s) at hand and are building a team around them, including Campbell Brown, well-known for her role heading news at Meta.  

Working with: Soon to be announced.

Questions still to answer:

  • How their measurement tool stacks up against security incumbents such as Cloudflare and Akamai.

  • Whether scraping equals using content and the relative value of each.

  • Whether LLMs will actually pay in this way.

Who do they want to hear from? Any news organisation. 

Contact: team@tollbit.com.

More information: https://www.tollbit.com.

ProRata.ai

TLDR: Building a tool that enables attribution and monetisation for content owners. This is based on what is being used in output rather than what is being scraped. 

The big picture: ProRata.ai’s view is generic deals are too wide and won’t accurately allow for attribution. The value isn’t necessarily about what is being scraped; it’s what users see and are therefore paying for (directly or through advertising). They are building a chatbot to trial the technology, but the real play is licensing their (as yet unseen) technology to LLMs to get fair representation and revenue based on output, not input. 

The deal: ProRata.ai takes 50% of revenues (this may change as their model builds out and if they can make the technology licensing deal work). 

The people: Heavy hitter CEO Bill Gross and media and tech veteran Annelie Jansen mean they understand both sides of the coin. If they can replicate Bill’s success with the invention of pay-per-click advertising, they will come out glowing. 

Working with: The Financial Times, The Atlantic, Axel Springer, The Guardian, Time, Fortune, plus other rights holders, including Universal Music Group and authors Tony Robbins, Seth Godin, and Scott Galloway.

Questions still to answer:

  • The technology has not yet been released for beta testing (but is due any day).

  • Who gets the coveted attribution on stories all news outlets are covering in a similar way.

  • Whether the LLMs will adopt the model and, if so, how payment will work.

Who do they want to hear from? Large English language publishers and consortiums of niche or non-English titles interested in using their tech. 

Contact: annelies@prorata.ai.

More information: www.prorata.com

Human Native

TLDR: Building a marketplace of content between LLMs and content owners with a premise that clean data is worth a vast amount more than scraped “dirty” data. 

The big picture: The two biggest cost centres for LLM are compute power and engineers. “Dirty” data needs some kind of verification to be cleaned, which uses a vast amount of compute power before it can be used. And as demand for engineers who can scrape has rocketed, they are now hugely expensive. The premise of the Human Native marketplace is that these costs can be decreased for LLMs and funnelled into licensing large, verifiable data sets. 

The deal: The content rights holder sets the price (with some advice if desired) and a 10% fee is levied to this when sold. 

The people: Ex-Google executives that have proven success at buildings and monetising, including the former head of partnerships for news Madhav Chinnappa. They err on the tech side but with a solid understanding of media. 

Questions still to answer: 

  • The value of pooled content and what size is “big enough.”

  • Whether rights can be segmented by use case, e.g., excluding licensing of data from competitors.  

  • Ongoing data management. 

Working with: Not public information. But our understanding is that it’s mostly the UK media, many international agencies, broadcasters, and book publishers. 

Who do they want to hear from? Anyone looking to monetise big data sets. Ideally, niche data that is not widely available on the Internet. They are open to conversations, but realistically much of this is demand driven at the moment so you need to have clear details of exactly what you can offer. 

Contact: madhav@humannative.ai.

More information: https://www.humannative.ai.  

And a bonus company I discovered just as this newsletter came out:

Scalepost.ai

In their words: Our platform connects publishers and AI/LLMs, facilitating smooth communication and collaboration. Streamlined agreements ensure a win-win for both parties.

According to their Web site, they manage the contracts with the publishers. On the surface, this looks like Human Native as above. But Perplexity uses Scalepost to manage partnerships, which means there is an ongoing data feed, not single sets of content. According to the publisher programme of Perplexity, they also provide AI analytics. 

More information: https://www.scalepost.ai

If you’d like to subscribe to my bi-weekly newsletter, INMA members can do so here.

About Jodie Hopperton

By continuing to browse or by clicking “ACCEPT,” you agree to the storing of cookies on your device to enhance your site experience. To learn more about how we use cookies, please see our privacy policy.
x

I ACCEPT