Data & ethics
What we will and will not do with the records we read
The archives we depend on are maintained for researchers and are sometimes approached by people whose interest is not research. Their custodians take a risk when they grant access. This page sets out the commitments we hold ourselves to, in specific terms rather than general ones, so that our conduct can be checked against them.
Core commitments
Research purpose only
We use archive data solely to produce public written research on removal requests and online speech. We do not use it for commercial purposes, for monitoring on anyone's behalf, or for any operational purpose. If our purpose ever changes, we will seek fresh permission rather than continue under an old one.
We respect redaction
Where an archive has withheld or truncated information, we treat that decision as final. We make no attempt to reverse redactions, correlate records to re-identify redacted parties, or reconstruct withheld URLs from other sources.
No republication of personal data
We publish aggregate figures and derived tables. We do not reproduce the names, addresses, contact details, or other personal information of senders, targets, or third parties appearing in records, including where an archive's researcher-level access would allow it.
No bulk redistribution
We do not republish bulk copies of the underlying records and we do not build a public mirror, search interface, or substitute for any archive we draw on. Verification of our work needs our parameters and our code, not a copy of someone else's database.
Where real people are involved
Our subject is state conduct, and our unit of analysis is the pattern rather than the person. Government bodies, courts, agencies, and companies are named freely: they act in an official capacity and accountability for official acts is the point of the research.
Private individuals are treated differently. We name a private individual only where they have already identified themselves publicly in connection with the matter and there is a clear public interest in doing so. Where naming would expose someone to legal or physical risk without a corresponding public interest, we describe the pattern without the name, even at some cost to the vividness of the finding. Journalists and civil society figures who are already public actors in a dispute are the ordinary exception, and even then we consider what our publication adds to their exposure.
Before publishing anything concerning an identifiable person, we consider whether the finding can be made without them, whether publication could contribute to harm, and whether they should have an opportunity to respond first. Where the answer to the last question is yes, we approach them before publication, not after.
Handling, storage, and retention
- Access control
- Extracted data is held only on encrypted devices belonging to members actively working on the study concerned. API credentials are never committed to source control, never shared outside the collective, and never embedded in published code or query strings.
- Minimisation
- We retrieve the fields our research questions require rather than everything available, and we discard fields we do not need at the point of extraction rather than storing them in case they become useful.
- Retention
- Working extracts are retained while a study is active and for a defined period afterwards to allow verification, then deleted. Published aggregates, parameters, and code are retained indefinitely, since they are the record of what we did.
- Breach
- If credentials or working data were ever exposed, we would notify the archive concerned promptly, revoke and replace credentials, and publish a note describing what happened.
- This website
- This site sets no cookies, loads no fonts, scripts, or images from third parties, runs no JavaScript, and carries no analytics. We do not know who visits it. For a project studying surveillance of speech, that seemed like the minimum consistent position.
Being a good guest on someone else's infrastructure
Research archives are run on limited budgets, and a careless researcher degrades them for everyone. Our technical practice reflects that: at least one second between requests, no parallel workers, compressed responses, narrow date-windowed queries instead of sweeping ones, and back-off rather than retry storms when a server signals that we are asking too fast.
Our requests identify us. Every call carries a descriptive user agent naming this collective with a contact address, so that an operator seeing unusual traffic knows immediately who to contact. We would rather receive an email asking us to slow down than be blocked as an anonymous nuisance.
Compliance with the Lumen API Terms of Use
Our access to the Lumen Database, should it be granted, is governed by Lumen's API Terms of Use. We have read them. Our specific undertakings:
- Research purpose
- We use the API only to produce public written research output, which is the purpose the terms permit.
- Honest identification
- We describe ourselves accurately in our application and on this site: an unincorporated, non-commercial research collective, with no claimed institutional affiliation and no publication record as yet. We would rather be assessed correctly than favourably.
- Required attribution
- The notice the terms require appears in the footer of every page of this site and will appear in every output that draws on Lumen data.
- Rate and stability
- We stay within documented rate limits and design our extraction so that it cannot degrade Lumen's service for other users.
- No substitute service
- We will not replicate or attempt to replace Lumen's own interface or user experience, and we will not sell, lease, sublicense, or share our credentials.
- Lawful use of linked content
- Where records contain links to material, our use of anything we retrieve stays within non-infringing and fair use, and we comply with applicable law.
- No implied endorsement
- We state clearly that Lumen has not endorsed, certified, or reviewed our work.
This product uses the Lumen API but is not endorsed or certified by Lumen.
Corrections and right of reply
Anyone who believes we have published something inaccurate about them, their organisation, or their agency may write to us and expect a substantive reply. Where we are wrong, we correct the text in place, mark the correction with its date, and say what changed. Corrections are listed on the publications page, never made silently.
We do not remove published research because a subject would prefer it unpublished. We do correct errors of fact, and we will publish a response from a party who disagrees with our interpretation alongside our own, provided it is signed.