Cashmere, APS strike AI discovery deal for physics research

Cashmere and the American Physical Society have partnered to make APS’s physics journals discoverable to agentic AI tools used by libraries and industry research teams. The deal gives APS control over access and usage while preventing its content from being used to train large language models.

Why it matters: - Academic libraries and industry R&D teams are using AI agents more often to search and synthesize scientific literature. - The partnership gives APS a controlled way to expose its physics research to that workflow without surrendering rights to AI platforms. - The model could influence how scholarly publishers license content for agentic AI discovery.

What happened: - Cashmere and the American Physical Society announced a partnership to bring APS physics research into agentic AI applications for discovery and inference. - The arrangement is designed so APS defines and controls the terms. - The announcement was made Oct. 26, 2026. - APS represents more than 50,000 physicists worldwide.

The details: - APS was founded in 1899. - APS publishes 19 journals, including Physical Review Letters, Physical Review X and Reviews of Modern Physics. - The journals form a peer-reviewed record of physics research used by academic, institutional and industry audiences. - Cashmere’s infrastructure lets APS license its catalogue for agentic discovery and inference while retaining control over which content is available, on which platforms and under what terms. - APS says the underlying research will not be used to train large language models. - Content will be held in Cashmere’s air-gapped infrastructure, which the company says is SOC 2, ISO 27001 and GDPR compliant. - Cashmere’s business intelligence layer provides usage insight into how APS journals are discovered and used by AI agents. - APS said that visibility can inform publishing strategy and support the long-term sustainability of physics publishing.

Between the lines: - The deal reflects a broader push by publishers to keep copyright control as AI becomes a new gateway to research discovery. - Cashmere is positioning compliance and usage tracking as the tradeoff for publishers that want AI distribution without open-ended model training. - For APS, the partnership offers a way to extend reach into AI-driven discovery while keeping authors and subscribers under a controlled licensing framework.

What’s next: - APS and Cashmere plan to expand access to APS journals through agentic AI tools used by libraries, researchers and industry customers. - The two organizations are also positioning the partnership as a template for other scholarly publishers. - Cashmere says it will continue to provide publishers with controlled inference, discovery tools and analytics.

The bottom line: - APS is betting that physics research can gain more visibility in AI tools without giving up control of its content, licensing terms or training rights.