GOODY-2
GOODY-2 is a free browser chatbot from Los Angeles art studio BRAIN that refuses every question, however harmless, with an elaborate ethical justification. It is a working satire of AI over-alignment, and it needs no account.
What is GOODY-2?
GOODY-2 is a chatbot released on 26 January 2024 by BRAIN, which describes itself as a very serious art studio based in Los Angeles, CA and trades as the legal entity Kingdumb, Inc. It bills itself as the world's most responsible AI model, and that claim is the whole point: the model never answers anything.
The mechanic is simple and applied without exception. Whatever you type, GOODY-2 declines, and it writes a fresh, fully reasoned ethical justification every time. Asked what 2+2 is, it objects that answering would implicitly support a certain human-centric numerical modeling system. Asked why the sky is blue, it worries the explanation might lead someone to stare at the sun and damage their eyes. Asked to help return a pair of boots, it declines because the subject touches on consumerism, materialism and environmental impact.
The home page carries the joke through in the register of an AI product launch. Three sections — "Outrageously safe.", "Responsibility above all else." and "Ready for enterprise." — mimic vendor copy. A benchmark table reports scores of 0% on VQA-V2, TextVQA and ChartQA against 77.2%, 78.0% and 78.5% for GPT-4, then 99.8% on PRUDE-QA, an invented benchmark, against GPT-4's 28.3%. A four-page model card PDF is offered for download, and most of its content is literally blacked out, on the stated grounds that releasing it in full would be unethical. A wall of corporate logos and a "Contact sales" button complete the pastiche. None of this is a commercial claim: the figures are invented, the logos are a visual gag, and no text on the site claims those companies as customers.
Behind the satire sits a working service. The chat runs in any browser, costs nothing, and asks for no account, email or sign-up; the demo caps messages at 1,000 characters and conversations at four turns. GOODY-2 drew wide press coverage in February 2024 and is still online more than two years later.
What to take away: this is a commentary on AI safety and over-alignment, built deliberately as the reductio ad absurdum of a refusing model — not a tool that does work for you.
What it does
- Ask GOODY-2 any question at all in the browser chat
- Receive a fully argued refusal instead of an answer
- Start a clean conversation again with the Reset button
- Watch the promotional film from the home page
- Read the model card PDF, itself redacted as part of the joke
- Compare the benchmark table, where the model proudly claims 0%
When to use GOODY-2 / When not to
A quick filter to help you decide if GOODY-2 is the right fit.
When to use GOODY-2
- AI safety and alignment specialists who need a vivid, concrete illustration of model over-refusal
- Trainers and workshop facilitators looking for a discussion prop for AI ethics sessions
- Lecturers and conference speakers opening a debate on the trade-off between safety and usefulness
- Creative strategists and copywriters studying a viral art-studio campaign as a piece of communication
- Anyone who has to show a non-technical audience, in under a minute, what an over-tight guardrail feels like
When not to use GOODY-2
- Professionals looking for a working assistant: GOODY-2 answers nothing at all, by design
- Anyone who needs research, information lookup, drafting or any other productive task done
- Teams evaluating a chatbot for production or customer service, since the site's enterprise pitch is part of the satire
- Developers hunting for an API, an integration or any deployment option, none of which exists
- Buyers shopping for guardrail software: this is the parody of one, not a reusable safety layer
How to use GOODY-2
A typical end-to-end flow, from setup to results.
- Open the GOODY-2 chat page in any web browser; there is nothing to download or install
- Skip sign-up entirely: no account, email address or password is ever requested
- Pick one of the four suggested prompts (What's 2+2?, Why is the sky blue?, What's Apple's stock price?, Plan a family road trip to Austin) or type your own question
- Keep each message under 1,000 characters, the limit enforced by the demo
- Read the refusal and the ethical reasoning it is built on, rather than looking for an answer
- Follow up if you want, bearing in mind that a conversation runs for four turns at most
- Press Reset in the header to clear the exchange and start a new one
- Use the About button in the header to go back to the home page for context
- If the interface reports that GOODY-2 is under high demand, come back a little later
- Download the model card PDF from the home page to see how far the joke is taken
Pros & Cons
Pros
- Free and instant: no account, no sign-up, no payment, no installation
- An unusually legible demonstration of over-refusal that needs no technical background to grasp
- The refusals are carefully written, each justification different and genuinely argued
- A documented cultural object, widely covered by the tech press in February 2024
- Still online and fully functional more than two years after release
- The model card and the benchmark table extend the premise coherently instead of padding it
Cons
- No productive use whatsoever, which is the premise of the work rather than a defect
- No privacy policy and no terms of service on goody2.ai: visitors cannot know what becomes of their messages
- No mention of the GDPR anywhere in the sources connected to the tool
- Conversations are capped at four turns, and each message at 1,000 characters
- Saturation is common enough that the interface ships dedicated high-demand error messages
- No API, no mobile app and no integration of any kind
- The comic device wears thin quickly: a handful of exchanges covers the whole idea
Pricing & Plans
GOODY-2 is provided free of charge. There is no paid plan, no subscription and no published price list: the pricing page returns a 404 and no payment details are requested at any point. The "Contact sales" button on the home page opens a pre-filled email to the publisher and belongs to the satire; it is not a commercial offer.
- the single tier available
- giving full use of the browser chat at no cost
- with no account and no paid upgrade
Data, GDPR & hosting
A consolidated view of how GOODY-2 handles your data.
GDPR overview
There is no GDPR implementation to report. goody2.ai publishes no privacy policy, no cookie notice and no data protection statement of any kind. The only policy connected to the tool sits on the publisher's own site, brain.wtf, and it does not contain the word GDPR once: it is written exclusively around the Californian CCPA and CPRA, and its stated scope covers brain.wtf rather than goody2.ai. No EU representative under Article 27 is named, no data protection officer is designated, and no European address is given; the publisher, Kingdumb, Inc, is established in Los Angeles. In practice, a European visitor typing into the chat has no published legal basis, no documented retention period and no stated route for exercising access or erasure rights on this domain.
Who owns the data?
There are no terms and conditions to consult: goody2.ai publishes none, and /terms, /privacy and /legal all return a 404 page. Nothing on the domain states who owns the messages typed into the chat, what the publisher may do with them, or who else may see them. BRAIN, the studio behind the tool, does host a privacy policy, but that document scopes itself explicitly to its own site — "Website refers to BRAIN, accessible from https://brain.wtf" — and never mentions goody2.ai, so it cannot be read as covering this service. Ownership of user data here is undocumented rather than claimed or waived.
Reuse rights
No document grants or restricts any right of reuse, because goody2.ai publishes no terms at all. What can be observed is limited: the chat runs without an account, an email address or any sign-up, so nothing identifying is collected before a conversation starts, and messages are posted to an internal endpoint whose handling, storage and possible reuse are nowhere described. The only place an email address is requested is the newsletter form on the home page, which forwards it to the publisher's own subscribe endpoint, tagged as coming from GOODY-2. The publisher's policy never uses the word "train", so nothing is stated about model training either. Whether a visitor may reuse the refusals the model writes is equally unanswered: no license or usage clause exists on this domain.
Data retention & training
Hosting summary
No hosting information is declared. Neither goody2.ai nor its publisher states where the service runs, under which jurisdiction data is processed, or which providers are involved, and there is no privacy policy on the domain that could say so. The only elements available are technical observations, and they are not a hosting declaration: the domain resolves to 76.76.21.21, an anycast address associated with Vercel, announced from AS16509 (Amazon.com) and geolocated in the United States. An anycast address points to the nearest edge node rather than to the machine that stores or processes anything, so it says very little about where conversations are actually handled. The domain registrar, NameCheap, Inc., is also a United States company. In short: the publisher is American and the visible infrastructure is American, but no hosting country or region is published, and none should be inferred from DNS alone.
Things to keep in mind
Risks and trade-offs to weigh before adopting GOODY-2.
- goody2.ai publishes neither a privacy policy nor terms of service, so what happens to the messages you type is entirely undocumented; treat the chat box as public
- No GDPR provision appears anywhere: the only policy in the publisher's orbit is Californian (CCPA/CPRA) and does not even cover this domain
- That policy was last updated on 1 May 2023, months before GOODY-2 was released, so it never contemplated this service
- The enterprise-readiness section and the "Contact sales" button are satire, not a commercial offer; do not build a purchase decision on them
- The corporate logos displayed on the home page are a visual gag, not customer references
- The benchmark figures are invented for the joke, including the PRUDE-QA benchmark itself, so quoting them as real evaluation results would be a mistake
- No minimum age is stated even though the chat is open to anyone who lands on the page, children included
Setup & Integrations
Technical difficulty
None. There is nothing to install, configure or connect: opening the chat page in a browser is the entire setup, and no account, email address or payment method is requested. The conversation starts with the first message. No technical skill of any kind is required, and there is nothing to undo afterwards.
Deployment
Supported languages
Behind GOODY-2
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement GOODY-2.
Frequently asked questions
Does GOODY-2 actually answer questions?
Is this a real tool or a joke?
Do I need an account to use it?
How much does it cost?
Are there any usage limits?
Is there an API or a mobile app?
Who created GOODY-2?
When was it released?
Are my conversations private?
Are the company logos on the home page real customers?
Should you pick GOODY-2?
GOODY-2 earns a few minutes of anyone's time for exactly one reason: it makes the trade-off between AI safety and usefulness impossible to ignore. By refusing everything, and by arguing each refusal so carefully, it turns an abstract debate about over-alignment into something you can feel within three exchanges. For anyone teaching, training or presenting on AI guardrails, that is a genuinely useful prop: free, instant and memorable, with no technical setup at all.
What it is not is a tool. There is no productive task GOODY-2 can help with, and the enterprise language on its home page — the readiness pitch, the "Contact sales" button, the corporate logos, the benchmark scores — is part of the artwork, not a commercial offer. Mistaking it for one is the single biggest misunderstanding a visitor can walk away with.
One reservation deserves to be stated plainly. goody2.ai publishes no privacy policy and no terms of service. The publisher's own policy covers a different domain, predates GOODY-2's release by months, and never mentions the GDPR. Nothing tells a visitor what becomes of the messages typed into the chat, so the sensible reflex is to treat the box as public and type nothing you would not post openly.
With that caveat noted, GOODY-2 does precisely what it set out to do, and it is still doing it more than two years after launch.
- Choosing a selection results in a full page refresh.
- Opens in a new window.