OpenAI ordered to delete ChatGPT over false death claims
OpenAI has been issued a Cease and Desist letter over “defamatory” statements allegedly made about a privacy advocate, whom the artificial intelligence falsely claimed to be dead.
Image by Shutterstock.
OpenAI, the startup behind ChatGPT, has been issued a Cease and Desist letter over “defamatory” statements allegedly made about Alexander Hanff, a privacy advocate, whom the artificial intelligence falsely claimed to be dead.
Hanff, a privacy technologist behind Europe’s General Data Protection Regulation (GDPR), issued the letter aimed at ChatGPT’s creators after realizing that the ChatGPT service told people he had died – generating fake URLs to ‘news’ stories it claimed as sources.
“Given that you had no legal basis to process my personal data for the training of your GPT large-language model, and that the data your model is producing is inaccurate, defamatory and harmful – I hereby demand that OpenAI delete the entire GPT-3.5 and GPT-4 models and all training data that contains my personal data,” said the letter, posted by Hanff on Linkedin.
The behavior that the privacy advocate describes broadly falls under a recent concept dubbed AI hallucinations, which occur when a response from an AI model looks convincing but has no factual basis in the real world. It can manifest, for example, as software code that doesn’t work, or a research document with made-up references.
According to Hanff, also known as That One Privacy Guy, it’s impossible to deconstruct the model in a way that would omit specific data about Hanff’s persona. He contends that OpenAI’s inability to do so leaves him vulnerable to defamation by the large-language program.
Moreover, the letter also said that OpenAI’s use of data constitutes questionable practice, as the model seems to use everything it finds online about a given subject indiscriminately – without giving any apparent consideration as to whether or not the information it is processing is true.
“That the data is available online is not a license to use it for commercial activities and does not provide an exemption to the requirements of Article 5 of the General Data Protection Regulation (GDPR), which requires that all processing of my personal data must be fair, accurate, lawful and meet obligations of transparency and data minimization,” Hanff’s letter said.
The letter ended with giving OpenAI, a startup Microsoft has thrown billions of dollars at, 30 days to cease using Hanff’s data before the company faces legal action.
“This is not a drill,” Hanff’s letter warned.
Earlier this month, Italy was one of the first countries to ban ChatGPT due to a data leak involving users’ chat messages and payment card data. However, OpenAI could relaunch in the European country if the company meets a set of demands by April 30.
Spain’s data protection agency has asked the European Union’s privacy watchdog to evaluate concerns surrounding ChatGPT. In response, the European Data Protection Board has already created a task force to keep the chatbot – and any of its AI successors – in check.