Skip to main content

Unfortunately we don't fully support your browser. If you have the option to, please upgrade to a newer version or use Mozilla Firefox, Microsoft Edge, Google Chrome, or Safari 14 or newer. If you are unable to, and need support, please send us your feedback.

Elsevier
Publish with us

Why are internal privacy and security critical to successful AI in R&D?

What to look for in a responsible, secure AI solution, and why you should involve IT and security stakeholders as soon as possible.

What are the risks of adopting AI tools for R&D?

If not monitored, AI tools in R&D introduce privacy and security risks that do not exist in conventional software: specifically, the risk of research data entered as AI prompts being used to train future models. These concerns need to be addressed at the outset, and organizations need to ensure key questions are answered early in the process to ensure their propriety information remains secure. In addition, if these questions are answered early, R&D teams can avoid lengthy delays caused by internal privacy and security assessments.

What are the privacy and security questions that shape every AI evaluation in R&D?

Below are some key questions that should be asked early in the process when exploring AI solutions for R&D. Always consult with internal AI and security teams for a detailed list of questions that are most relevant to your organization.

“Will this tool see or track my research activity?”

A lack of data privacy risks exposure of your R&D team’s research intent, strategy and any potential discoveries. This could be done through search tracking, individual user-level activity records or surveillance by internal or external actors.

What should you look for?

Ensure users' searches are private and only accessible to them, with fully encrypted, private and secure usage.

“Are AI models being trained with my data or search activity?”

Many generalist AI tools are trained through user activity – this includes inputs, queries and other activity. While this is common, this is a non-starter for an R&D team looking to use AI to maintain a competitive advantage.

What should you look for?

Ask whether sessions are secure, with no training based on user activity. Searches, history and uploaded documents should not be used to train any models.

“What happens to the content I upload into the tool?”

R&D teams need to be able to upload documents and other proprietary data to make the most of any research tool. However, they need to be sure this content remains secure – there should be no fears of storage, reuse or leakage, ensuring your data remains your data.

What should you look for?

All uploaded documents should be encrypted, session-specific and only visible to the uploader. Once users delete an uploaded document, the vendor does not retain it in any form.

“Can anyone else, internal or external, see my work on the platform?”

R&D teams need to be able to research with freedom, without exposing their methodology to other parties, such as colleagues, IT teams and external vendors with access to systems.

What should you look for?

Data should be user-isolated, not visible to other users or vendors and only sharable if the user chooses to share it.

“How reliable is the underlying content fueling the AI tool’s responses?”

Most AI shows sources, but in order to see real impact in R&D, researchers need confidence in where the responses are coming from. This can mean all the difference when it comes to risk management and helps avoid making key innovation decisions based on bad data.

What should you look for?

R&D teams should prefer AI tools that use only licensed, vetted, peer-reviewed sources. Responses should include traceable references so researchers can know for sure where the outputs came from.

Why you should start AI conversations with IT and security now

The sooner IT and security teams are brought in, the easier it will be to get these tools in the hands of your teams. If these stakeholders aren’t involved until too late, it often adds weeks or months to the procurement process – costing R&D teams valuable time where they could be reaping the benefits of AI-assisted research. By discussing your needs with stakeholders sooner, this will position them as partners in the quest to find the right solution, not roadblocks.

Right away, you should work to identify the following:

  • Who internally should be engaged in these discussions?

  • What are your internal AI governance requirements?

  • What security and IT reviews should you expect

  • What are the risks of waiting too long to involve internal stakeholders?

Elsevier’s approach to AI privacy and security

At Elsevier, we understand that trust is earned. We are dedicated to protecting your data, respecting your privacy and upholding the highest standards of security and compliance. Explore how our policies and practices support responsible innovation and safeguard your information. 

LeapSpace addresses many of these privacy and security concerns, with the goal of creating a tool that R&D teams can use every day, with confidence that their data and search activity is protected. 

  • In LeapSpace, your activity is private, secure and encrypted. We never use your data to train any LLMs. 

  • LeapSpace runs on a private infrastructure, not a shared AI ecosystem. This ensures high privacy and security standards, with secure user sessions and strict data governance. 

  • LeapSpace’s privacy and security guardrails are designed to help protect customer IP and support data privacy. 

  • LeapSpace’s closed, curated data ecosystem pulls from only trusted scientific content – not unvetted web sources. 

  • LeapSpace supports secure, scalable use across enterprise research environments. R&D teams gain deep, critical insights from trusted scientific content, while helping organizations keep their IP and data to themselves. 

Learn more about Elsevier’s approach to safeguarding user information.

Explore Elsevier's advanced data solutions supporting AI in R&D, including datasets, connectors and more

Young man working as software developer, coding on computer in modern office