Article

Using AI within your business – key considerations

29 October 2024

AI data processor chip

AI and its uses

Artificial Intelligence (“AI”) is a field of computer science that creates machines or software that can perform tasks that would typically require human intelligence. For example, AI can be trained to understand speech – like Siri or Alexa – play games like chess, or recommend films – as with Netflix.

AI uses data on a mass scale, following set instructions, or algorithms, to learn from experiences. The more data an AI system gets, the better it becomes at understanding patterns and making decisions.

Data use in AI

AI programmes can process data of various types, including images, audio and video, and structured data.

There are three main categories of data within the AI lifecycle: training data, input data and output data.

  • ‘Training data’ refers to the data used to train the AI system – for example, publicly available data or privately sourced or purchased data-sets
  • Input data’ refers to information contained in prompts and instructions provided to the AI to run a scenario and generate an output. For example, Spotify uses a subscriber’s listening history – the “input data” – to generate personalised playlists – the “output data”
  • Output data’ refers to data contained in the outputs generated by the AI, for example a written note, image or audio file.

When using AI within your business, it’s important to understand where the training data is coming from and whether any of your input or output data may be used by the AI for further machine learning and to improve the tool for the benefit of others.

All of the above categories of data could contain protected or higher risk data such as sensitive personal data and works subject to third party intellectual property rights.

Key considerations for users of AI

Various rights can attach to data, such as copyright, confidentiality and database rights. If the AI provider has not obtained the appropriate consents and licenses, there could be potential for legal claims, such as breaches of copyright, confidentiality or personal data.

Whilst any liability should primarily sit with the AI provider, if you utilise AI, it’s advisable to know where the training data was from, how it was obtained and what consents are in place. Whilst the UK government intends to produce a code of practice to allow for some uses of copyrighted works within AI models, this has not yet been implemented and it is very much a moving area.

You may want to seek assurances that the AI provider will verify the accuracy and quality of the data, but in practice, most software providers (AI included) will disclaim any responsibility for accuracy of results generated. It is also worth asking the AI provider how the model prevents biases.

Related articles

Resource
3 January 2025 2 minute read

HCR Flex

Read more
View All