What should I look out for with AI in research & science?

Never use AI as a source: every reference must be traceable to the original article, as fabricated or wrongly attributed citations are the most common error. Unpublished data, research proposals and peer-review manuscripts are confidential or commercially sensitive, so do not enter them into a consumer account without a data processing agreement, and check whether the provider stores your input or uses it for training (publishers usually prohibit this explicitly for peer review). If you work with personal data from research participants, the GDPR and your own ethics review apply: running models locally or using an EU-hosted tool with data minimisation is then the safe route. Finally, document your search strategy and your AI use: journals and funders increasingly ask for this, and a review must remain reproducible.

Home / Safety / Claude for research & science

Anthropic

How safe is Claude for research & science?

GDPR profile for research & scienceStrong

Claude has a well-documented GDPR profile and, with the right business settings, is genuinely usable in research & science.

Where does Claude stand in research & science?

Shared spot 1 of 30 options for research & science (professional tools and general assistants) on GDPR profile. 5 options have a Strong GDPR profile, just like Claude. Each tile is an option we track for this field; click for the check.

Strong5

Good6

Limited8

Unknown4

Caution7

Published safeguards

✓EU hosting possibleYes

✓No training on your data (business)Yes

–Self-hostable (on-premise)Not known

Data is stored in the US by default; EU data residency is possible via AWS Bedrock or Google Vertex AI (EU region). Business tiers and the API do not train on your input; consumer tiers do, unless you opt out.

How to use Claude safely

Use the business variant and sign a data processing agreement (DPA).
Do not enter unnecessary personal data into prompts; anonymise where possible.
Enable EU data residency where the provider offers it.
Keep a professional in the loop: never use AI as a source: every reference must be traceable to the original article, as invented or misattributed citations are the most common error.

Stronger or equally strong in research & science

CradleCradle (Amsterdam, the Netherlands and Zürich, Switzerland)

ASReview LABUniversiteit Utrecht (the Netherlands)

KeeniousKeenious (Norway)

ScholarcyEverway Opco Limited / Scholarcy (Antrim, United Kingdom)

SciSpacePubGenius Inc. (Milpitas, United States)

Latest developments

9 November 2026Claude Code stores approved commands containing passwords in settings.local.json

If you choose "Yes, and don't ask again" in Claude Code, the tool saves the command verbatim in .claude/settings.local.json, including passwords or API keys. Apiiro found real secrets in 1,122 of 104,185 public copies on GitHub, including 379 database passwords. Keep the file out of git, use variables for secrets, and replace any keys that have already leaked. source

November 2 2026Message threads (beta) fall outside Zero Data Retention

With message threads in the Claude API, Anthropic stores the conversation, the system prompt and the tools on the server, encrypted in memory and not on disk, until up to one hour after the last use. The feature does not qualify for Zero Data Retention; your other ZDR agreements still apply. If you work with personal data under ZDR, only enable threads if your DPIA covers it. source

26 October 2026Swedish government office blocks the browser version on work computers

From 26 October, the Swedish government office blocks ChatGPT and Claude on work computers because there is no formal contract and there are information security risks; Microsoft Copilot remains permitted. Business use therefore requires a contract with data processing terms. source

Questions

How safe is Claude for research & science?: Claude has a strong published GDPR profile and, with the right business settings, is well suited to research & science. EU options, no training on your data and/or self-hostable.
Does Claude train on my data?: Data is stored in the US by default; EU data residency is possible via AWS Bedrock or Google Vertex AI (EU region). Business tiers and the API do not train on your input; consumer tiers do, unless you opt out.
What should I look out for with AI in research & science?: Never use AI as a source: every reference must be traceable to the original article, as fabricated or wrongly attributed citations are the most common error. Unpublished data, research proposals and peer-review manuscripts are confidential or commercially sensitive, so do not enter them into a consumer account without a data processing agreement, and check whether the provider stores your input or uses it for training (publishers usually prohibit this explicitly for peer review). If you work with personal data from research participants, the GDPR and your own ethics review apply: running models locally or using an EU-hosted tool with data minimisation is then the safe route. Finally, document your search strategy and your AI use: journals and funders increasingly ask for this, and a review must remain reproducible.

Source documents

platform.claude.com claude.com

Everything about Claude Best AI for research & science

Summarised by the editorial team from provider documentation · not legal advice

Last checked: 9 October 2026 · How we check

Next stepAll 30 options for research & scienceProfessional tools and general assistants side by side, on GDPR profile, price and quality.

Every Friday: what changed in AI for research & science

New models, prices, oversight and tools, with sources and what it means for you. One email a week, unsubscribe with one click.

View a sample

Your fieldEmail address

We only use your address for this newsletter. You confirm first by email.