The Information Commissioner’s Office (ICO) recently launched a consultation series on how data protection laws should apply to the development and use of generative AI models (“Gen AI”). In the coming months, the ICO will publish further views on how to interpret specific requirements of UK GDPR and Part 2 of the DPA 2018 in relation to Gen AI. This first part of the consultation focusses on whether it is lawful to train Gen AI on personal data scraped from the web. The consultation seeks feedback from stakeholders with an interest in Gen AI.

As outlined by the ICO, web scraping will involve the collection and processing of personal data, which may not have been placed online directly by the data subjects themselves. To comply with the UK GDPR, Gen AI developers would need to ensure there is a valid lawful basis for their processing under UK GDPR, as well as comply with the relevant information requirements pertaining to indirect personal data collection.

For the first part of the consultation series, the ICO published a policy position on the lawful basis for training Gen AI models on web-scraped data which can be found here. More specifically, this consultation focusses on the ‘legitimate interest’ lawful basis under art. 6(1)(f) UK GDPR and the ‘three-part’ test that a data controller must pass to meet the legitimate interest basis (a so-called Legitimate Interest Assessment). The ICO has considered various actions that Gen AI developers could take to meet this three-part legitimate interest test to guarantee that the collection of training data through web scraping, i.e. processing of data, is complaint with the principles of UK GDPR. The ICO would now like to hear from relevant stakeholders on their view of the proposed regulatory approach and the impact this would have on their organisation. A link to the survey can be found here.

The deadline to submit a response is 1 March 2024.