Scientists worry AI is stealing their ideas

Researchers are increasingly concerned that large language models are incorporating unpublished or proprietary ideas from their grant proposals and private manuscripts into future outputs. This practice threatens the integrity of the peer-review process and creates potential conflicts regarding intellectual property and academic credit. As AI integration grows in scientific workflows, scholars worry that these systems may inadvertently expose sensitive data to broader training sets, undermining the confidentiality required for new discoveries.