- OpenAI no longer relies on API customer data to train its multilingual models and ChatGPT, CEO Sam Altman told CNBC.
- The company quietly updated its terms of service in March to reflect the changes.
- Altman said customers “clearly want us not to train with data.”
OpenAI hasn’t been training big language models for AI, such as GPT, on paid customer data for “a while,” CEO Sam Altman told CNBC on Friday.
“Customers clearly want us not to train with data, so we changed plans,” Altman told CNBC’s Andrew Ross Sorkin.
OpenAI’s terms of service were quietly updated on March 1st, breaking records for the Internet Archive’s Wayback Machine show. “We haven’t done any training on API data at all. We haven’t in a while,” Altman told CNBC. An API, or application programming interface, is a framework that allows a customer to plug directly into his OpenAI’s software.
OpenAI’s enterprise customers such as Microsoft, Salesforce, and Snapchat are more likely to take advantage of OpenAI’s API capabilities.
However, OpenAI’s new privacy and data protections apply only to customers using the company’s API services. The company’s updated terms of service state, “We may use content from services other than our API. This includes, for example, text that employees enter into our extremely popular chatbot ChatGPT. It’s possible: Amazon recently warned employees not to share sensitive information with ChatGPT for fear that sensitive information would appear in responses.
This shift comes as the industry grapples with the prospect of replacing human-created material with large language models.
The Writers Guild of America, for example, went on strike Tuesday after negotiations between the Guild and film studios broke down. rice field.
Executives are equally concerned about the intellectual property impact of ChatGPT and similar programs. Barry Diller, the entertainment mogul and chairman of the IAC, has suggested that the media company could take the matter to court and creative he could sue an AI company over the use of his content.
