Should we impose a confidentiality agreement, akin to attorney-client privilege, on the sensitive data we disclose to AI?
Sam Altman discussed the possibility in a recent interview, speculating that society might determine that there is some sort of AI privilege.
There are legal and medical privileges when you speak with a lawyer or doctor, according to Altman. As of right now, conversing with an AI lacks that concept, but perhaps it should.
During a discussion about their new AI health company, Thrive AI Health, with media mogul Arianna Huffington and OpenAI CEO, the subject came up. The product of the company is an AI health coach that monitors user health data and offers tailored advice on nutrition, exercise, and sleep habits.
The regulation of data storage and sharing has gained traction as businesses have further integrated AI systems and products.
Sensitive health information on a patient cannot be disclosed by a doctor without the patient’s consent due to laws like HIPAA. In order to promote better care, the agreement is intended to make patients feel at ease enough to be honest with their physicians.
However, some patients still find it difficult to communicate with physicians or to seek medical care; Altman claimed that this was one of the reasons he decided to get engaged with Thrive AI.
Altman was shocked by how many people were prepared to provide information with large language models, or LLMs, which are AI systems that underpin chatbots such as ChatGPT and Google’s Gemini. He stated that he had read Reddit discussions about people who had success telling LLMs things they were uncomfortable revealing with others.
Although the product’s design is still being finalized by Thrive AI, Huffington stated that she envisions it being “available through every possible mode,” including workplace platforms.
This raises questions regarding regulations and data storage. Large IT corporations have already been sued on the grounds that they trained their AI models on unlicensed content. Businesses might utilize health data to train LLMs since it is among the most valuable and confidential information that individuals own.
It would be crucial to explain to individuals how data privacy operates, Altman told. “But from what we’ve seen, most people get this,” Altman continued.
The goal of Thrive AI Health, which was introduced last week by OpenAI’s Startup Fund and Thrive Global, was to address “growing health inequities” and “democratize access to expert-level health coaching” through the use of AI.
