twitter_dataset_release_rules.txt

Public Release Rules for Twitter/X-Derived Research Datasets (1.7M posts)

1) Bottom line (what we may publish)
- Do not publish tweet text or JSON objects. Public bulk redistribution of X/Twitter content is restricted to Post IDs (tweet IDs) and User IDs only.
- Share IDs + our own derived dataset, i.e., MMT_LID dataset, which can be shared with text.

2) X Developer Policy — redistribution rules (binding)
- IDs-only in bulk. If we provide X content to third parties (e.g., downloadable datasets), we may only distribute Post IDs / DM IDs / User IDs.
- Cap per recipient: We may not distribute > 1,500,000 Post IDs to any one entity within any 30-day period without written permission from X. As a result, we have a reduced version of file: MMT_distributable_without_tweets, which is under 1.5M posts (i.e., in the ballpark of 1.35M) 
- Small, manual exception: We may provide up to 500 public Post Objects and/or User Objects per person per day via non-automated means (e.g., a small CSV/PDF). This does not cover our full dataset.
- Academic carve-out (IDs only): We may distribute IDs for non-commercial research on behalf of an academic institution if the research is approved by X in writing or has peer review/validation; only the necessary number of IDs may be used.

Key numeric limits (from X policy)
- Bulk IDs per entity: ≤ 1,500,000 Post IDs within 30 days, unless X approves more (in writing)
- Manual sharing of objects: ≤ 500 public Post/User Objects per person per day (non-automated). So here we can send text also, which is under 500 posts. 

3) Data protection & privacy laws (we must still comply)
- EU GDPR / EDPB: Posts can be personal data. Pseudonymisation reduces risk but does not remove GDPR obligations; additional measures are required to prevent re-attribution.
- California CCPA/CPRA: “Personal information” excludes certain publicly available information, including information a business reasonably believes was lawfully made available to the general public. This does not override X’s contract terms and does not permit bulk redistribution of tweet text.

What we will release:
1) Post IDs (tweet IDs) and, if necessary, User IDs.
2) Our derived artifacts: task labels, splits, and aggregated features that do not reproduce X content (no text, no media, no per-post geodata extraction).

What we will not release:
- Tweet text, media, or full tweet/user JSON in bulk.
- Standalone geolocation traces or any linking of handles to off-platform identifiers without opt-in consent.

Compliance checklist:
- Dataset contents: Confirm the archive contains only IDs + our derived fields; purge any text/media/JSON remnants.
- README & license: State that the dataset contains only IDs and must be hydrated via the X API by requesters; instruct requesters to honor deletions/privatizations and 24-hour compliance.
- Takedown process: Publish a contact email; commit to removing or modifying entries within 24 hours upon verified request from X or the post owner.
- Access gating: Because of the 1.5M IDs / 30-day per-entity cap, provide staged downloads or host via a controlled repository that meters access per entity.
- Display examples: If we show illustrative posts on the project page, use official embeds or follow the Display Requirements.
- Geo/identity safeguards: Ensure no standalone geo traces or off-platform identity linking without consent.
- Privacy law note: Flag that downstream users hydrating IDs may process personal data and must comply with GDPR/CPRA; include a pointer to GDPR pseudonymisation guidance.

Statement for dataset page:
“This release contains only Post IDs/User IDs and our derived labels/metadata. We do not distribute tweet text, media, or full tweet/user JSON. Requesters must hydrate IDs via the X API and must honor deletions, privatizations, and other changes; any valid takedown request will be processed within 24 hours. Distribution is subject to the X Developer Policy, including the 1,500,000-IDs limit, and to applicable privacy laws (e.g., GDPR/CPRA). As a result, we have a reduced version of the file: MMT_distributable_without_tweets, which is under 1.5M posts (i.e., in the ballpark of 1.35M).”