🧠

NeuralNews

AI Alignment & Machine Psychology
Legal Notice & Privacy

Who runs this site, and what happens to what you type

NeuralNews is a private research project. It has no accounts, no advertising and no analytics — but it does send what you type to language-model providers, and one of its features publishes to a page everyone can read. Both are spelled out below.

Before you type a topic. Whatever you enter in the Explorer, the Trend Radar or the Roundtable is transmitted to a language-model provider in order to answer it. Treat the input boxes the way you would treat a search engine: no unannounced plans, no contract internals, no client or customer names, no details about identifiable people. And note that the Roundtable offers to save a debate — if you use it, your topic and the resulting closing statement become publicly readable in the Salon gallery.

Provider

Dr. Pantaleon Fassbender
16751 NE 5th Street
Williston, FL 32696
United States of America

E-mail: pantaleonfassbender@gmail.com

This website is a private research project of a natural person resident in the United States. It is not operated on behalf of an institution, an employer or a publisher, there is no company behind it, and it carries neither advertising nor sponsorship. Responsible for the content: Dr. Pantaleon Fassbender, address as above.

Controller and applicable law

The site is operated and hosted in the United States. It is reachable from the European Economic Area, so for visitors there the General Data Protection Regulation applies by virtue of Art. 3(2) GDPR; the controller within the meaning of Art. 4(7) GDPR is the provider named above. Processing therefore takes place outside the EEA.

What this site is, technically

Eight static pages and a set of serverless functions. There is no user account, no login, no contact form and no newsletter. The site sets no cookies. It runs no analytics, no tag manager, no advertising and no session-recording service, and it does not track you across time or across other websites.

It also loads nothing from anyone else's server — no fonts, no scripts, no images, no stylesheets. The typefaces are served from this site. Opening a page contacts exactly one host: the one in your address bar. You leave it only when you click through to a source.

One qualification, because it is true and the paragraph above would be misleading without it:

  • Session storage. The Explorer keeps the result of a search in your browser's sessionStorage so that switching tabs does not re-run the query. It never leaves your device and your browser discards it when you close the tab.

Server logs

Hosting is by Netlify. Like any web server, its infrastructure logs the requests it serves — typically IP address, timestamp, requested URL, HTTP status, bytes transferred, user agent and referrer. This is technically unavoidable and is the only routine server-side collection; it serves operation and security, it is not analysed by the provider of this site, and the retention period follows Netlify's own. Legal basis: Art. 6(1)(f) GDPR.

What each tool sends, and where

Nothing leaves your device until you ask for something. What happens then differs by tool:

  • Explorer. Your query is used to fetch matching articles from the public Google News RSS index — fetched by this site's server, not by your browser. The query and the retrieved article headlines and summaries are then sent to Google's Gemini model, through Netlify's AI Gateway, to extract concepts and relationships.
  • Trend Radar. A term you enter is run against the Google News index server-side. The resulting score is cached server-side for roughly an hour (the top-ten board, roughly three hours), and that cache is shared — a term you scored may be served from cache to another visitor. The cache holds the term and its score, not anything about you.
  • Narrative Lens. Takes no input from you. It classifies a rolling sample of recent headlines with Gemini and caches the dashboard for about three hours.
  • Innovation Index. Read-only. The daily scores are produced by an unattended evening job, not by your visit.
  • Roundtable. Your topic is sent to Anthropic's Claude models: first to a screening step, then — if it passes — to the six panellists and the facilitator, one phase at a time, with the accumulating transcript. The debate is held in your browser's memory and is not stored anywhere unless you choose to save it.
  • Salon. Reading the gallery sends nothing. Saving does — see below.

Legal basis for all of the above: Art. 6(1)(b) and (f) GDPR — performing the operation you asked for.

The public gallery

This is the one place where something you did is kept and shown to other people, so it is worth reading twice. If you save a debate you convened, the following is written to this site's database and becomes publicly readable by anyone in the Salon gallery, with a timestamp, for as long as the site exists: the topic you entered, the facilitator's closing statement, its key findings, and the identifiers of the readings it selected. Nothing about you is stored alongside it — no IP address, no identifier, no account, because there are no accounts.

Because there is no account, there is also no self-service way to delete an entry. If you want one removed, write to the address above naming it and it will be deleted. Separately, an unattended job adds one entry each evening from a machine-chosen trending topic; those are not connected to any visitor.

Recipients

Netlify, Inc. as host and as operator of the AI Gateway, the cache and the database; Anthropic PBC as the provider of the Claude models; Google LLC as the provider of the Gemini models. All are in the United States. Google receives only what a tool sends it to answer your request — it is not a recipient of anything simply because you opened a page. There is no other recipient, and nothing is sold or passed on for marketing purposes.

Retention

  • Server logs — per Netlify's retention periods; not analysed here.
  • Trend and narrative caches — about one to three hours, then overwritten.
  • Explorer session cache — in your browser only, discarded when the tab closes.
  • Gallery entries and daily index rows — kept indefinitely, because they are the archive; gallery entries are deleted on request.

Your rights

You have the right of access (Art. 15 GDPR), rectification (Art. 16), erasure (Art. 17), restriction (Art. 18), data portability (Art. 20) and objection (Art. 21), and under Art. 77 the right to lodge a complaint with a supervisory authority. Requests go to the address above. The answer will usually be short, because apart from the server logs — and anything you deliberately saved to the gallery — nothing about you is held here.

No representative in the Union under Art. 27 GDPR has been designated. The provider relies on the exception in Art. 27(2)(a): occasional processing, no large-scale processing of special categories of data, and no likely risk to the rights and freedoms of natural persons. Should that assessment cease to hold, a representative will be designated and named here.

Notice for California residents

Under the California Online Privacy Protection Act (Cal. Bus. & Prof. Code §§ 22575–22579): the categories collected are network-activity information in the form of the server logs described above. Name, postal address, e-mail address and telephone number are not collected — the site has no field in which to enter them. Text you type into a tool is transmitted to the model provider to answer it, and is not retained by this site unless you deliberately save a debate to the public gallery. Recipients are named above; no sale or sharing for marketing takes place. There are no user accounts and therefore no stored profile to review or amend; reloading the page discards everything entered.

Do Not Track: this site does not track visitors over time or across third-party websites, so it does not change its behaviour in response to a Do-Not-Track signal — there is no tracking to switch off. Material changes are published on this page with a new date; there is no mailing list for individual notification.

Children

This site addresses adults in a professional or scholarly context. It is not directed at children and does not knowingly collect data from them.

External links

The site links to news articles, papers and other external material. Those pages are the responsibility of their respective operators. Links were checked when added; continuous monitoring without specific cause is not practicable. Links will be removed promptly if something objectionable is brought to notice.

No warranty

Offered free of charge and without warranty. Everything on this site is produced by language models reading public sources, and any of it can be wrong. In particular: the Trend Radar measures news coverage as a proxy for virality, not actual searches; the Narrative Lens classifies a bounded sample of headlines, not whole articles; the Innovation Index is one model's judgement, published with its sources so it can be checked rather than trusted; and a Roundtable debate is argument, not fact — six personas are made to disagree on purpose. The closing readings are resolved from a hand-verified library so they cannot be invented, but the note explaining each one is model-written. Verify anything you intend to rely on.

Status: 19 August 2026. If this statement and the behaviour of the site ever diverge, the statement is the thing that is wrong, and it will be corrected — the description follows the code, not the other way round.