Data Policy

Last updated: 28 September 2026. Version 1.0. Read with our Privacy Policy.

This policy explains how Implausible handles experiment and research data: how it is protected, who can use it, how long we keep it, and what control you have. The honesty of the data is the product. So is the way we look after it.

1. Principles

• Minimum data. We collect what an experiment needs, and nothing else.
• Separation. Names and contact details are kept out of research data.
• Pre-registration. Hypotheses, methods and stopping rules are registered before data is analysed.
• No selling. Participant data is never sold, and never used for advertising.
• Evidence over belief. We publish null and negative results, not just interesting ones.

2. How your data is layered

Your experiment data is stored under a random account ID, not your name or email. That lets us send your own results back to you. When our research team looks across participants, each person appears only as a research code made with a keyed hash. The key is held separately, with restricted access. Every export is given new random IDs that cannot be linked to your account, to other exports, or to the research codes. Exports contain no free text, no date of birth and no exact times, and are only produced when at least 10 consenting participants are included.

This is pseudonymisation, not anonymisation. Pseudonymised data is still personal data under UK law, and we protect it as such. We only call data "anonymous" once it has been aggregated or processed so that no one can reasonably identify you, taking account of all means likely to be used.

3. How experiments protect data integrity

• Targets are generated with publicly verifiable randomness (drand) and sealed with a cryptographic hash before each session.
• Seals and timestamps are logged so results cannot be altered after the fact.
• Warm-up trials are excluded from analysis.
• Patterns suggesting gaming, automation or multiple accounts are flagged and excluded.

4. The Metamodel and AI

The Metamodel is a statistical and machine-learning system. It learns from experiment and state data to:

• give you personalised feedback on your prediction performance
• find patterns across participants who have given consent
• test hypotheses registered in advance

Population-level patterns and community priors use only consenting participants, and only groups of at least 30 sessions.

We do not use your identifiable data to train third-party AI models. Where a third-party AI provider processes data for us, such as the AI model that blind-judges ARV impressions, it acts as our processor, receives no name, email or account ID, and may not use the data to train its own models.

Model outputs are estimates, not facts. A pattern in the model is a hypothesis to test, not a finding.

5. Consent

Consent is given on the sign-up screen, in its own unticked box, separate from agreeing to the Terms. It reads:

I give my explicit consent for Implausible to use my experiment data, including my wellbeing and belief answers, as described in the Privacy Policy and Data Policy. I can withdraw at any time in Settings.

It is a condition of using the platform, because experiments cannot run without this data. We record when you consented and to which version of the wording.

Where the law allows, your consent covers the field of research into human prediction and intuition, rather than one study only. We will ask again before we collect a new type of data (such as heart rate or other wearable data), use your data for a purpose outside that field, or share it in any form other than anonymised exports.

6. Withdrawing

You can withdraw consent at any time in Settings → Privacy, or by email. When you do:

• we stop collecting experiment data from you immediately, and experiments are paused
• your data is excluded from all population-level research straight away
• your account stays open, and you can still sign in and download your data
• if consent stays off for 30 days, we delete your experiment data
• data already part of a completed analysis, a published result or an anonymised export cannot be withdrawn from it, but it cannot be linked back to you

If you switch consent back on within 30 days, your history is kept and experiments resume.

To close your account completely, use Settings → Privacy → Delete my account. We act on deletion requests within one month.

7. Sharing with research partners

We share research data only with universities, research institutes and independent researchers who:

• have a registered study with a stated purpose
• sign a data sharing agreement that bans re-identification, onward sharing and commercial use without our written agreement
• apply appropriate security and delete or return data at the end of the study
• receive the minimum fields their study needs, as an anonymised export

Re-identifying de-identified personal data without consent is a criminal offence in the UK under section 171 of the Data Protection Act 2018. Our agreements say so.

We share aggregated, anonymous statistics in publications, reports and public dashboards.

8. Commercial use

Implausible also runs consultancy work and an internal, proprietary fund. Personal data from participants is never shared with either, and is never used to make individual trading decisions. Any commercial use of research findings relies only on aggregated, anonymous results.

9. Retention

10. Security

• Encryption in transit (TLS) and at rest.
• Role-based access, multi-factor authentication for staff, and access logging.
• Keyed research codes, with the key held separately.
• Row-level security in the database, so each user can only reach their own records.
• Processor contracts with every provider that touches personal data.
• A breach response plan. We report notifiable breaches to the ICO within 72 hours, and tell affected people without undue delay when there is a high risk to them.

11. Accountability

We keep a record of processing, a data protection impact assessment for the experiment platform and the Metamodel, and a legitimate interests assessment for each purpose that relies on it. We review this policy at least once a year and whenever we launch a new experiment type.

Questions: privacy@implausible.io.