1. Who operates scime
scime is a research workspace made by Latentic Labs. This notice describes account information, research content, connected services, and optional data sharing. A separately hosted deployment may have its own operator and data-handling arrangements.
2. Workspace information
scime processes your name, email, sign-in method, verification status, and profile image when provided. It stores research content needed for the features you use, such as project files, references, ideas, AI instructions and drafts, collaboration records, connected profiles, imported publications, and accessible OpenReview discussions.
News language, region, topic, and excluded-source preferences personalize the feed. Region preferences describe interests and do not request your precise location.
3. Google and OpenReview
Google sign-in uses identity information returned by Google, including your identifier, verified email, name, and image when available. Google Scholar is a separate connection using the public profile URL you provide. Google sign-in does not itself grant access to Gmail or Drive. Scholar exports download your saved publication list.
Connecting OpenReview sends your credentials and any verification code to OpenReview. scime does not save your OpenReview password; it stores an encrypted session token. It imports submissions and discussions accessible to your account. Disconnecting removes the local connection, imported OpenReview records, drafts, and statistics contribution, without deleting originals on OpenReview.
4. AI and external services
When you request AI assistance, relevant inputs are sent to the AI you chose in Settings: scime's own AI model, run by Latentic Labs on its server in the Republic of Korea (the default), or the provider of an API key you added, such as OpenAI, Anthropic or another OpenAI-compatible service. Depending on the task, these can include manuscript text, selected reviews, references, instructions, and your current draft. scime counts your daily AI requests on its own model and does not use your inputs to train AI models. API keys you add are stored encrypted and are used only for your own requests. Workspace drafts and related records are retained to support your work.
Only share material you are authorized to send to the selected provider. Provider processing and retention terms also apply; scime does not promise that an external provider never retains inputs or uses them for other purposes.
The service providers listed below process information needed for their functions when used. External links lead to independently operated services. Loading external images can cause your browser to contact their hosts.
5. Service providers
These companies process information for scime when the related feature is used. Several are based outside the Republic of Korea, mainly in the United States, so information may be processed in other countries.
Amazon Web Services (hosting): servers and databases that store your account and workspace data.
OpenAI, Anthropic and other AI providers (AI): only when you add your own API key for them, the inputs of AI features you request, as described above.
Google (sign-in): identity information when you sign in with Google.
Resend (email): your email address and the message content for verification, invitations and account notices.
GitHub and Zotero (optional connections): only when you connect them, using the access you grant, to sync project files or import references.
Reference lookups: Semantic Scholar, Crossref, arXiv, DBLP and OpenAlex receive the titles, identifiers or search terms needed to verify references. These lookups do not include your account information.
scime atlas (Latentic Labs): research ideas and abstracts are compared with the atlas paper index to find related work.
6. Optional score statistics
Statistics contribution requires a separate affirmative opt-in. Connecting or syncing OpenReview does not opt you in. Only author submissions are eligible; reviewer-only assignments are excluded.
The private snapshot contains paper and review identifiers for coauthor deduplication, venue, score field, scale, score, and update information. It excludes manuscript text, review text, and reviewer identities. Histograms show each paper's average accessible review score, separated by conference, year, track, rating field, and scale.
Only users with active consent and a current score contribution can view community statistics. Groups appear from one paper and one author profile. A histogram may therefore reveal an individual paper's average even though identifiers are not displayed. Grouping does not guarantee anonymity.
Withdraw through My submissions → Community statistics to remove your contribution and lock statistics access. The same paper can remain through a coauthor's independent consent. Snapshots stop contributing after 30 days unless refreshed; expiry alone does not delete a stored snapshot. Withdrawal, disconnection, or account deletion removes your contribution. Previously viewed or downloaded results cannot be recalled.
7. Cookies and browser storage
Essential cookies support sign-in and authentication security, including temporary OAuth state. Browser storage holds interface preferences, guest news preferences, and saved Library items. Browser-saved items may remain on a shared device after sign-out.
Clearing browser storage may remove saved items and preferences or sign you out. The current application does not install third-party advertising or analytics cookies. This does not control external websites you visit.
One first-party cookie (scime-src) remembers for up to 30 days which link or site first brought you here: its utm tags or the referring site, or, when there is neither, which public page you opened first (for example the home page or a guide; never a project, an invitation or anything you typed). If you create an account in that time, this is saved with the account so we can count where sign-ups come from; administrators see it, and it is deleted with the account. The public pages (the home page, pricing, guides, tools, the sign-in page and the pages about OpenAI Prism and Overleaf) also report a page view to scime atlas (atlas.scime.ai, also run by Latentic Labs): the page, the referring site and the utm source. Pages of your workspace are never reported. That count uses no cookie, stores no IP address and tells visitors apart only within a day.
8. Retention and your controls
Use Settings to manage your account and AI usage, workspace pages to manage connections, and project controls to manage content. Collaborators access content according to permissions you grant. Export important materials before deleting a shared project or account.
Workspace information is retained to provide your account and features until removed through available controls or handled following a request. Account deletion is available in Settings. When an account is deleted we keep one record without your name, email, sign-up source or content: when the account was created, last used and deleted, how you signed in, how many projects and labs it had, and the reason you chose, if any, so we can see why people leave. Account deletion does not clear browser-local data, exported copies, or original records held by connected services. Operational records and backups may be handled separately from active workspace data; contact us about deletion scope and timing.
Contact Latentic Labs for access, correction, export, deletion, or privacy requests not covered by the interface. Rights and necessary exceptions depend on applicable law. Account ownership may need to be verified before acting on a request.
9. Changes to this notice
The date above identifies this version. Service or data-practice changes may require an updated notice. This notice does not replace the separate opt-in for OpenReview score sharing.
Questions or requests
Contact Latentic Labs at contact@scime.ai. Please do not include passwords, API keys, or confidential manuscripts in your message.