Anthropic, a generative AI startup, has made a commitment to not utilize customer data for training its large language models (LLMs), as detailed in a recent update to the terms of service for Claude, its developer-focused product. Furthermore, the company has pledged to offer support in cases of copyright disputes.
Led by ex-OpenAI researchers, Anthropic updated its commercial terms of service, effective from January 2024, to clarify its stance. The updated terms ensure that the outputs generated from Anthropic’s AI models belong to its commercial customers, with the company stating that it does not plan to claim any rights to customer content under these terms.
This move follows similar pledges made by other tech giants like OpenAI, Microsoft, and Google in the latter half of 2023. These companies promised support for customers facing legal challenges due to copyright claims linked to the use of their technologies.
Anthropic’s revised terms offer customers protection against copyright infringement claims that may arise from authorized usage of its services or outputs. The company highlights that users of Claude will benefit from increased protection and a more user-friendly API.
In addition to this, Anthropic has committed to covering the costs of any approved settlements or court judgments related to copyright infringements caused by its AI. This applies to customers using Claude via its API and those employing Claude through Bedrock, Amazon’s generative AI development suite. The terms specifically mention that Anthropic does not aim to acquire any rights in the content provided by customers, nor does it imply rights to any party's content or intellectual property.
LLMs like Anthropic's Claude, along with others like ChatGPT-4 and Llama 2, rely on extensive text data for training. The quality of an LLM is largely dependent on the diversity and comprehensiveness of this training data, which enhances accuracy and contextual understanding by learning various language patterns, styles, and new information.
In a related development, Universal Music Group filed a lawsuit against Anthropic AI in October 2023, alleging copyright infringement of numerous works, including lyrics of many songs under its control. In a separate case, author Julian Sancton is suing OpenAI and Microsoft for allegedly using his work without authorization to train AI models, including ChatGPT.
