[reposted from my LinkedIn account]
I ask this only partly in jest.
The story of how ChatGPT refused to acknowledge the existence of “David Mayer” and some others, perhaps (probably?) because people with those names had exercised their erasure rights (such as the right at Articles 17 of the GDPR and the UK GDPR), raises the interesting question: if a sufficient number of people made such requests, would the LLM begin to fail?
If so, a further question of rights arises. If I, Jon Baines, exercise my erasure right against ChatGPT (or another platform/LLM), and it suppresses any processing of the words “Jon Baines”, what effect might that have on my namesake Jon Baines, and his travel company? Or Jon the Ocean Specialist working on the Ocean Watch program?
Because the words “Jon Baines”, in isolation are not my personal data. In isolation, they do not relate to me. A crude response to an erasure request, just as with any of the other crude approaches which AI is capable of (for instance in relation to accuracy), runs the risk of interfering with others’ rights, including rights to operate a business, our rights to freedom of expression.
I don’t have an answer, but this is just one extra point and possible flaw in AI which will no doubt play out over the coming years.
The views in this post (and indeed most posts on this blog) are my personal ones, and do not represent the views of any organisation I am involved with.
