Anthropic has agreed to maintain copyright protection measures in its AI systems as part of an agreement with music publishers. This development comes in response to a preliminary injunction motion filed in the Northern District of California.

Key points of the agreement:

• Anthropic will maintain existing filters on user query responses
• The company can improve and optimize these guardrails as long as their effectiveness isn't reduced
• New AI models and products must include similar copyright protections
• Publishers can notify Anthropic if they find their content being improperly used
• Anthropic must investigate and respond to publisher complaints promptly

This agreement partially resolves the dispute that began in October 2023 when eight music publishers sued Anthropic. The publishers sought an injunction in August 2024, claiming it was necessary to prevent copyright infringement. Anthropic had argued its use of copyrighted content for AI training qualified as "fair use" due to the transformed nature of the output.

The stipulation maintains that no party admits liability, fault, or wrongdoing. The music publishers' complaint regarding unauthorized use of lyrics for AI model training remains unresolved.

According to the agreement, if publishers identify instances where guardrails fail to prevent reproduction of their content, they must notify Anthropic in writing. Anthropic is then required to investigate and provide a detailed written response outlining their plan to address the issue or clearly state if they don't intend to make changes.

Related Articles

Previous Articles