AGI Scorecard中文
All sections
Legal

Privacy policy

Last updated: October 3, 2026

This page explains what data agiscorecard.com collects and how it is used. The short version: the site includes public research pages and optional interactive services, including discussions, AI tools and paid cloud workspaces. We do not sell your data.

Analytics

On public pages, Google Analytics starts by default unless you have opted out. Use the footer analytics toggle to disable or enable it. Existing explicit refusals, Do Not Track and Global Privacy Control are respected. The choice is stored in this browser. This optional channel receives the public page address and title, the referring origin and ordinary Google Analytics browser metadata; it does not receive tool inputs, uploaded files, URL queries or fragments. First-party counters remain separate. Public templates share this analytics channel; private account and delivery pages remain excluded.

Google Analytics 4

We use Google Analytics 4 to understand aggregate traffic — which pages are visited, roughly where visitors come from, and which site features are used. GA4 sets cookies and collects pseudonymous usage data (such as page views, approximate location at city level, and device type). We do not use this data to identify individuals. You can opt out with the Google Analytics opt-out browser add-on or by blocking analytics cookies in your browser.

First-party log (added August 5, 2026)

Alongside GA4, and not instead of it, the site keeps its own small log so that the public experiment results on /experiments can be reported from numbers we control rather than from a third-party dashboard. It is stored in a Cloudflare D1 database owned by this site and has two parts.

Page counts, recorded at the server. When this site's Cloudflare Worker serves a page, it adds one to a counter for that day, page path, referring site's hostname, country, campaign tag, and whether the request looked like a bot. That is the whole record: a tally, not a log of visits. No JavaScript is involved, so — unlike the parts below — a content blocker does not switch it off.

Interaction events, recorded from the page. A short script mirrors the same interaction events GA4 already receives (for example: which subscribe button was clicked, which quiz answer was picked) to /api/e on this domain, with the page path, referring hostname, country, browser language and any campaign tag. Only a fixed list of event names is accepted; anything else is discarded on arrival. Blocking JavaScript, or any content blocker that blocks /api/e, prevents this part.

Sets no cookie and stores no identifier. Neither part stores a cookie, a session ID, a device fingerprint, an IP address, or anything that links one visit to another. (Interactive services have separate storage described on this page; newsletter addresses are described below.) Referring sites are stored as a hostname only, never the full referring URL, which can carry query strings that identify people; of the query string on our own pages, only utm_source, utm_medium and utm_campaign are read. Rows cannot be traced back to a person, which also means we cannot single out and delete "your" rows — there is nothing there to identify.

Work Mentor practice

Work Mentor keeps fictional-case answers and practice history in local browser storage by default. Export a backup before clearing browser data. When you explicitly request AI coaching, this site and Cloudflare Workers AI process your question, fictional task and answers. The application does not save the question or model reply in its database. Do not submit personal or company secrets. Hashed daily network identifiers enforce the shared preview quota.

Optional usage measurement and explicitly submitted feedback use a pseudonymous browser identifier. The database stores allowed event or feedback categories, language and coarse source/campaign labels, not report answers, question text, raw referrer URLs or raw IP addresses. These records are retained for up to 90 days. Test events marked as QA are excluded. You can leave usage measurement unchecked. The separate site analytics choice still applies.

Cloud saving occurs only when you choose Save on the member page and requires the existing AGI membership. That service stores the backup you submit, its versions and payment/access records under its stated terms. Opening the member page alone does not upload the backup. Local stages and self-reported work application are not identity-verified credentials.

Newsletter

The email briefing is operated on beehiiv. If you subscribe, your email address is stored by beehiiv and used solely to send the newsletter; every issue includes a one-click unsubscribe link. See beehiiv's privacy policy for details of their processing.

Signing up on this site (added August 6, 2026)

Subscribe buttons used to send you to a form on beehiiv.com. You can now enter your address without leaving the page, so this site itself receives and stores your email address — that is a real change from the position described above, and interactive services have their own processing described on this page.

Stored with the address: which button you used and, where it applies, which prediction's alert you asked for (so a "tell me if this verdict flips" signup can actually be honoured), the page you signed up from, your browser language, a two-letter country code from Cloudflare's edge, the referring site's hostname, and any campaign tag in the link. No password, no account, no profile — there is nothing to log in to.

It is used to send you the briefing and the alerts you asked for, and for nothing else. It is not sold, rented, or shared with anyone beyond beehiiv, which sends the mail. Where beehiiv is connected, the address is passed to it immediately and you get a confirmation mail; where it is not, the address waits in our database until it is imported, and the page says so rather than promising a mail that will not arrive.

Getting it removed. Every issue has a one-click unsubscribe. To have the address deleted from this site's own database as well, reply to any issue and ask — deletion means deletion, including the signup record described above.

Portfolio market widgets

The portfolio tracker displays TradingView quotes and charts in cross-origin frames when the market section enters view. TradingView receives ordinary connection data, such as the IP address, browser information, selected symbol and public page origin, under its privacy policy. Our integration passes no research notes, calculator inputs, account keys or page query strings to these frames. Pause market data to remove the frames. Public performance-record checks read this site only and pause when hidden or offline.

Hosting

The site is served via Cloudflare, which may process standard technical request data (IP addresses, request logs) to deliver and secure the site, per Cloudflare's privacy policy.

Advertising

The site currently shows no advertising. If that changes, this policy will be updated first, and any advertising partner's required disclosures will be added here before ads are served.

Your rights

Depending on your jurisdiction (e.g. GDPR, CCPA), you may have rights to access or delete personal data. For newsletter data, use the unsubscribe link or contact beehiiv. GA4 data is pseudonymous; blocking cookies prevents collection going forward. The first-party log holds no identifier at all, so there is no record tied to you to retrieve or erase. Blocking JavaScript or /api/e stops the interaction events being written; the server-side page counter is a per-day tally with nothing personal in it and is not blockable, which is stated here rather than glossed over.

Contact

Questions about this policy: reply to any issue of the email briefing.

Community accounts and discussions

Free AGI community accounts use a private login key and public display name. Approved contributions are public; submissions awaiting review are private. Followed topics, moderation and minimal visit statistics stay in the AGI database. Community profiles do not grant paid workspace access. Read the community privacy details and deletion controls / 社区隐私与删除说明.

AI infrastructure research

Watchlists and notes stay in the current tab until you choose to save. Local copies use localStorage and can be read by others using this browser. Only records explicitly saved in the member page are uploaded to the same-site cloud workspace, under its storage and expiry terms. AI review sends only the selected company’s public financial facts to Cloudflare Workers AI, never your notes. Abuse prevention stores a keyed hash of the request IP, not the plaintext IP. Counters have a twenty-four-hour window and expired counters are removed in bounded batches on later requests; without later requests they may remain longer.