Anthropic, Promise Not To Use Customer Data For Big Language Model Training

Anthropic, an artificial intelligence (AI) startup that specializes in the genrative AI, (photo: x @AnthropicaI)

JAKARTA - Anthropic, an artificial intelligence (AI) startup that specializes in AI-generatives, promises not to use customer data for large language model (LLM) training, according to updates to the commercial service provisions of developer Claude. The company is also committed to supporting users in copyright disputes.

Anthropic, led by ex-OpenAI researchers, revised its commercial service provisions to clarify its position. From January 2024, the change states that Anthropic commercial subscribers also have all the results of using its AI model. The company "does not plan to acquire any rights to Customer Content under this provision."

OpenAI, Microsoft, and Google pledged to support customers facing legal problems due to copyright claims related to their use of technology in the second half of 2023.

Anthropic is also committed to the updated commercial service provisions to protect customers from copyright infringement demands emerging from legitimate use of the company's services or proceeds.

"Restrictions will now enjoy greater protection and calm of mind as they build with Claude, as well as an easier-to-use API," said Anthropic.

As part of his legal protection pledge, Anthropic said he would pay for any settlement or approved decisions resulting from his AI violations. This provision applies to APIcus customers Claude and those using Claude through Bedrock, Amazon's generative AI development suite.

In connection with copyright infringement, Universal Music Group sued Anthropic AI in October 2023 for copyright infringement of "a large number of copyright-protected works - including lyrics for various musical compositions" under the ownership or control of publishers.

Meanwhile, author Julian Sancton sued OpenAI and Microsoft for alleged use of the work of the non-fiction author without permission to train AI models, including ChatGPT.