AI models are hungry for your data, so don’t just give it away

Taking stock of your proprietary data as an intangible asset is the first step in monetising it or realising additional value to your data.

Artificial intelligence (AI) developers compete these days for high quality data as they separate themselves from the pack. Often this means tapping into and training their models on good quality proprietary data.

But where will AI developers and deployers find high quality data?

The answer: You, if you let them.

This article discusses the value of your organisation’s data, why it matters in AI deployments, why everyone wants your data and concludes with five recommendations for you if you are considering adopting an AI solution for your enterprise.

What is “Proprietary” Data? – Intellectual property rights in data
When it comes to data, the question of whether there are legal property rights can be a complicated one.

Firstly, it depends on what the “data” is. Mere publicly available factual information per se might not attract legal property rights, but a secret formula or process might be protected by law as a trade secret, with common law confidentiality obligations attaching to anyone to whom the information is given in circumstances imparting an obligation of confidentiality. And companies pay good money for access for commercially valuable trade secrets.

Secondly, different countries may provide different treatment. For instance, Singapore does not have sui generis database rights, but the EU has a database directive that provides a form of legal protection over databases, which in turn can lead to value extraction from the right to

And we can also speak of “proprietary” data in a non-legal sense, as information that is privately controlled by a company or individual. “Proprietary” data, in this sense, tends to sit behind enterprise firewalls, or in systems sequestered securely from outsiders or unauthorised users.

Whether this is proprietary data in the legal sense or in the sense of privately controlled data, the value of such proprietary data has been accentuated in the AI race.

Why proprietary data?
Publicly available on the internet is a varied landscape data. Most of it is not curated, sorted, enriched or created in a way that truly captures value. Purposefully curated and processed data has far more value than incidentally collected raw unfiltered data.

“Value” also means different things to different enterprises. An e-commerce platform would have data users who need data for different reasons. Merchants on the platform need data to customise their offerings or prices. Suppliers need data to hedge demand and supply patterns at appropriate times of the day or year. Consumers who have shopping preferences may want personalised recommendations.

Data also needs to be enriched, and this often happens with the participation of subject matter experts who know their domains. Basic RAG systems can be oblivious to important context, flounder on complex queries, and occasionally give wrong information with the flair of an expert. Virtual experts might become a real offering soon but human subject matter experts are essential in curating, and supporting the fine tuning processes involved in using AI.

Take the example of GPT legal, a large language model (LLM), which makes legal research quicker and more efficient, the product of a collaboration between the IMDA and Singapore Academy of Law. A key feature of GPT Legal is the fine tuning of generic LLMs with legal data, with advanced retrieval capabilities and testing / refinement with subject matter experts. The subject matter experts are not necessarily bots – they’re lawyers who know their work.

The investments involved in the development of such purpose-specific and commercially valuable models or applications include man hours, financial resources, expenditure in tools and other resources.

Your company probably has Proprietary Data
It is a good bet that every enterprise has data of some value which it keeps confidential or protects from its competitors. It may be data that arises from the way in which you do business, or data points which are valuable to you and which you watch closely, or it may be insights you derive from that data.

Knowing whether this data has value or not involves conducting an assessment. Intellectual property assessments and stock taking – usually done by an intellectual property lawyer are key steps to undertake.

Competitive value can be found even in the edges of the most incidental information.

Everyone is inventing ways to get your data
To illustrate, let’s take the hypothetical example of a company in the ESG space which offers carbon footprint and data analysis of its customer’s organisation in the form of a report, for free.

To get this report, the customer does not pay a cent. Instead, the customer provides information via a guided questionnaire, giving profile data and information generally about its business. The report that comes out contains detailed analysis and reporting on the customer’s carbon footprint and recommendations for improving the environmental credentials.

But the ESG company is not a charity. As more companies take up their free offer of generating a report, the ESG company’s long game becomes clear. As more quality data it collects, it can develop newer and better AI models, and other AI solutions emerge – each trained on data that would otherwise have been hard to access.

When adopting AI, know your intellectual property and what the bargain is
With this in mind, it is good to understand that it is not just a matter of cybersecurity – it is a question of whether you are giving out your data for free, and when considering whether to take on an AI solution, it is good to consider some key points:

In what way is my “data” protected by intellectual property rights, and what commercial value is there? Are there trade secrets, patentable inventions, copyrighted works, and other forms of legal property in my data that I should  be aware of?

How will the AI solution use the data provided to it, for instance, will the AI be trained on my data?

Am I extracting value from my data with AI solutions in a way that enhances my data as an asset?

Look for the fine print, and if you cannot see anything there, ask and get assurances that your proprietary data is never used or shared for training without your consent.

And if I am consciously agreeing to allow an AI to use my data, what is the bargain that’s being made with the AI developer for this?

Taking stock of your proprietary data as an intangible asset is the first step in monetising it or realising additional value to your data.

The important thing is not to take the value of your data for granted. As the quest for higher quality data in the AI race shows, there’s value to be had in your data, and no one should give it away for nothing.