LinkedIn has suspended training of its AI models using UK user data after an intervention by the Information Commissioner’s Office.

Microsoft — LinkedIn’s owner and multibillion-dollar investor in OpenAI, maker of ChatGPT — is no stranger to data protection investigations. Just in the past year or so, there are many other examples:

• In June 2024, the ICO issued a statement that it was liaising with Microsoft over its Recall function. Some commentators have labelled Recall a form of uninstallable spyware. Others have raised concerns over security holes.
• In March 2024, the European Data Protection Supervisor ruled that the European Commission’s use of Microsoft 365 infringes data protection law for EU institutions and bodies.
• In November 2023 the Irish Data Protection Commissioner reprimanded Microsoft for failing to handle erasure requests in a compliant manner.
• In June 2023, Microsoft announced it was making provision for a fine of $425m by the Irish DPC. This fine was due to LinkedIn’s targeted advertising practices. It seems this ruling has not been finalised or made public yet.

As the race for commercially viable generative AI solutions continues, we will see this tension over use of personal data more frequently. After all, generative AI is only as good as the underlying datasets on which it has been trained. It’s not surprising AI providers are doing a data grab. But obtaining a lead in the market should not come at the expense of personal data.

What should you do?

If you are a provider of generative AI solutions, you need to check you have consent to use the datasets. Here, the ICO is investigating the use of personal data. But, as we have seen before, the dataset might also contain proprietary intellectual property rights or confidential business information. Or it might contain inaccurate data or even AI-generated output, thus creating an AI echo chamber. Your AI tool is only as good as you make it. If you get this wrong, your competitors will likely beat you anyway.

If you are a customer looking to use AI, you should do your research to understand what you’re getting. Remember, if you’re using mass-market AI, you’ll get it on the same terms as everyone else. This includes the usual unhelpful disclaimer that it is provided “as is”. If you’re taking all the risk, you are responsible for ensuring the output is reliable. If you can’t get better terms for more reliable AI – or at least for reliable and affordable AI – then you should implement safeguards to offset the risks. If your clients rely upon your faulty output generated by poor quality AI, you could face a legal claim.

If you need advice, contact me at +44 20 3824 9748 or fjennings@hcrlaw.com.