SciSpace Academy
Tutorial

Can You Trust AI with Your Research Data? Privacy and Safety

SciSpace AcademyEnglishResponsible AI

TL;DR

1-min read

This SciSpace tutorial addresses a common hesitation about AI tools: what happens to unpublished drafts, interview notes or confidential manuscripts once they are uploaded. It is aimed at researchers who want to understand how their documents are handled before using AI, and it separates common myths from facts about data use.

Key takeaways

  • The tutorial explains what happens to drafts, notes and papers after upload to SciSpace, including where they are stored and who can access them.
  • It addresses whether uploaded content is used to train AI models; for any tool, read the privacy policy and terms rather than assuming, and look for clear statements on training use, retention and deletion.
  • Safeguards such as encryption and secure storage are covered, along with SciSpace's approach to institutional and journal compliance and how sensitive content and academic integrity are handled.
  • Before uploading unpublished work, check your institution's and funder's rules as well as the tool's data policy.
  • As a general rule, avoid sharing identifiable participant information, health records, data under confidentiality or ethics agreements, and manuscripts you are reviewing for a journal.
  • When in doubt, anonymise material and consult your ethics board or data protection office, and recognise the limits of AI use in day-to-day research.
Apply it

To apply this, review what you plan to upload against your institution's guidance, anonymise sensitive material, and then use SciSpace for the tasks that fit those boundaries.

How to do it

  1. Learn what happens after upload

    Find out what happens to drafts, notes and papers once you upload them to SciSpace. This addresses the core AI research data privacy concern.

  2. Check storage and access

    Confirm where uploaded data is stored and who can access it. Knowing this helps you decide what to upload.

  3. Confirm whether content trains models

    Check whether content you upload is used to train AI models. This separates myth from fact before you share sensitive work.

  4. Review the safeguards in place

    Review how encryption, secure storage and compliance practices protect sensitive research content. This informs your institutional and journal obligations.

  5. Set your own safe-use habits

    Decide practical boundaries for what you upload and how you use AI day to day. Good habits keep your research safe.

Watch the tutorial44 min

Can You Trust AI with Your Research Data? SciSpace Privacy ExplainedPlay video: Can You Trust AI with Your Research Data? SciSpace Privacy Explained

Chapters

  1. Welcome and agenda
  2. What's at risk when you upload research data to AI tools?
  3. Data-driven AI vs privacy-first AI: what's the difference?
  4. Does SciSpace train AI models on your uploaded papers?
  5. What data does SciSpace collect, and why?
  6. In-transit vs at-rest encryption explained
  7. Is end-to-end encryption enough? What to look for in AI tools
  8. How to permanently delete your data: hard delete and GDPR rights
  9. Is SciSpace SOC 2 Type II compliant?
  10. Who can access the files you upload to SciSpace?
  11. Can other users or AI models see your papers? OpenAI and Claude via Azure and Bedrock
  12. Is it safe to upload unpublished research papers to an AI tool?
  13. When you delete a chat or file in SciSpace, is it really gone?
  14. Are uploaded papers encrypted during transfer and storage?
  15. How does SciSpace prevent data breaches?
  16. How PDFs are parsed in temporary, isolated environments
  17. Can you export or fully delete all your SciSpace data?
  18. Which AI models does SciSpace use, and can you choose them?
  19. Is AI safe for clinical and sensitive research data?
  20. Key takeaways on privacy, security and user control
  21. How AI guardrails block prompt injection and data leaks

Try this on SciSpace

Frequently asked questions

Is it safe to upload unpublished research to AI tools?

It depends on the tool's data policy. Before uploading, check where data is stored, whether it is encrypted, who can access it and whether it is used for model training. Also check your institution's and funder's rules, and avoid uploading identifiable participant data or confidential material unless the tool is explicitly approved for it.

Do AI tools use my documents to train their models?

Policies vary between providers, so read the privacy policy and terms of service rather than assuming. Look for clear statements on training use, data retention and deletion. This tutorial explains SciSpace's own practices on these points; for any tool, prefer providers that state their position plainly and let you remove your data.

What research data should I never share with an AI tool?

As a general rule, avoid sharing personally identifiable participant information, health records, data covered by confidentiality or ethics agreements, and manuscripts you are reviewing for a journal. When in doubt, anonymise the material, consult your ethics board or data protection office, and follow your institution's guidance on approved tools.

Related tutorials