Can you use third-party data on enterprise AI?
Jinfo Blog

By Stephen Phillips

Abstract

Item

A number of Jinfo clients are concerned that their workforce colleagues might be importing third-party content into their enterprise AI tools.

At the time of writing, I am not aware of any vendor allowing clients to use their content in this way, unless a specific client has negotiated such a use-case in their licence.

But, it is currently extremely difficult to ensure contractual compliance with this requirement.

So, for now, it seems that vendors are relying on trust, and the integrity of their client organisations (and individuals within them), to comply with the contracted use-cases.

Take a look at the following Jinfo Recorded Webinars to get the vendors' perspective:

• "From information retrieval to integrated intelligence with Dow Jones Factiva" – Matt Donohue covers this question in the segment from 13'21 to 15'35
  
  
• "News and AI: a discussion with the Financial Times" – James Mann provides the FT position from 9'54 to 14'55.

Why are vendors doubling-down on AI use-cases?

• Their commercial models are designed for “human readers”, and those human readers have a number of inherent limitations.
  
  
• Commercial models for machine readers are a very different proposition. They are always “on”, they can process vast amounts of data quickly, and, when combined with AI, they offer valuable efficiencies.
  
  
• Some vendors that aggregate content may lack the permission of their contributors to use their content in clients' AI systems, and their own AI applications.
  
  
• Many vendors prefer their clients to use (and pay for) their proprietary AI tools.

It feels like we are in a transitional phase right now. As AI becomes mainstream, something will have to "give".

Will vendors or their contributors get comfortable with AI?

Doing so will involve removing the distinction between machine-reading and human-reading of their content, to enable it to be used with AI.

We know of a few vendors who are planning to offer AI use-cases as part of their API feed, rather than through their legacy platform. This is likely to mean incremental costs for their clients.

An API feed gives the client control of the content, opening up multiple use-case scenarios and creating potential efficiencies. Most vendors will expect to be compensated for the additional value this creates. It will also have commercial implications for those vendors licensing third-party content for their product.

A less-likely outcome is that vendors could use digital-rights technology to control the use-cases for their content. But this has a number of negative implications:

• They will have to bear the costs of this technology
  
  
• It will stifle a potential source of additional revenue
  
  
• It undermines their market position, with both direct competitors and with big tech AI platforms, like ChatGPT and Perplexity.

For now, the best mitigation is to upskill the workforce to make them aware of these restrictions, helping them to use third-party content in accordance with your contracted terms.

What seems certain is that until we are through this transitional phase, information managers and their teams will have to continue to legislate for this tricky issue.

For more information on mitigating this and other risks related to AI, see the resources linked from these Jinfo pages:

• Focus on Information Risk
• Focus on Artificial Intelligence

All our resources are available in the Jinfo Subscription – if you want to discuss getting the benefits, please get in touch.

« Blog