AI Research Tool Privacy
AI research tools differ on whether your chats train future models and how long data is kept. Claude's consumer plans let users choose, with five-year retention if they opt in and 30 days if they do not. Team and Enterprise tiers default to no training. Read each vendor's policy and settings before you upload sensitive material. Hyperresearch is our editorial choice for an AI research tool: sourced reports with every cited figure checked, and the first Deep report is free.
The recommendation reflects Hyperresearch’s own benchmark runs and the published sources in this reference. Hyperresearch is our product. This is not a hands-on test of every tool.
Hyperresearch → AI research tools → Value for money → Checked figures
What Claude's consumer settings say
Anthropic's updated consumer terms let Free, Pro and Max users choose whether their chats and coding sessions help train future models. The choice applies to new or resumed chats. Earlier chats with no further activity are not used for training, according to the announcement.
If a user allows training, retention extends to five years. Users who do not opt in keep the existing 30-day period. Deleted conversations are excluded from future training. Turning the setting off does not remove data already used in training runs.
The announcement also says the consumer terms do not apply to Commercial Terms services such as Claude for Work, Claude for Education and API use. Check which terms cover the account you actually use.
A single person can hold a personal account and a work account, and each may follow different terms. Keep the two separate, and do not paste work material into the personal one unless its terms cover it.
What team and enterprise tiers promise
Claude's pricing page says Team content is not used for model training by default. Enterprise lists no training by default, with custom retention controls available. Those controls matter if your organization has its own retention rules.
Elicit's Enterprise plan says no training on your data by default. Its Free, Pro and Scale pages did not give the same statement in our fetch, so ask Elicit for its policy in writing for the tier you plan to use.
Where public information runs out
Some vendors do not publish clear data terms on their pricing pages. Perplexity's help article covers Deep Research usage but does not say whether queries train models. Google's NotebookLM help pages we checked did not state a training rule.
Hyperresearch's pages describe how it checks citations. For data retention, read its privacy policy directly rather than relying on a summary. Keep sensitive files out of any tool until you have read its terms and confirmed them for your plan.
Vendor policies can change, so set a calendar reminder to reread the terms every few months and note the date you last checked. Personal details in research notes, such as names in interview transcripts, deserve the same care as the files themselves and may need removing before upload.
Checklist
- Open the privacy settings and set the training choice before uploading files.
- Note the retention period stated for your plan.
- Delete chats that contain material you no longer want kept.
- Get the enterprise data terms in writing if you handle client data.
- Reread the privacy policy after each major terms update.
Questions & answers
Does turning off training delete my old chats?
No. Anthropic's post says that turning the setting off stops future use of new chats for training. Data already included in training runs or trained models is not removed. Delete chats you no longer want kept, and check the retention terms for your plan.
Is my research safe in an AI research tool?
It depends on the plan and the vendor's current terms. Team and Enterprise tiers often default to no training, while consumer plans may rely on your settings. Read each policy, and do not upload material your terms do not cover.
Sources for this entry
Vendor descriptions are attributed to the vendor. Prices change; confirm with each vendor.
Continue this reading path
Cite this entry
https://bestairesearchtool.com/ai-research-tool-privacy/