How Does the AI Safety Chatbot on AISafety.info Work?

Updated

AISafety.info hosts a chatbot called Stampy that answers questions about AI safety. You type a question into the chat input on the site, and Stampy responds. The site itself warns that Stampy "can be inaccurate" and tells users to "always verify its sources," so treat it as a fast starting point for orientation rather than a citable authority.

What Stampy is for

Stampy exists to give quick answers to AI safety questions without requiring you to read through the site's full article library first. The homepage presents it as the "Questions?" entry point, alongside the site's structured content:

  • Basic sections and Advanced sections of articles
  • Browse all articles and browse by category for self-directed reading
  • How Can I Help? and Get involved for people who want to contribute
  • Donate for financial support

So Stampy is one of several ways into the same material, not a replacement for it.

How to use it

  1. Find the chat input on the site (the homepage shows a prompt field with placeholder text like "Try asking me...").
  2. Type a question about AI safety in plain language.
  3. Read the answer, then follow up on any sources it cites.
  4. Cross-check anything important against the site's articles or other references.

The expected result is a short, conversational answer that points you toward relevant topics. The site's own disclaimer is the key operational detail: because Stampy can be wrong, verification is part of the workflow, not an optional extra.

What the site says about AI risk

Stampy's subject matter is framed by the site's core claims, which appear as the main headings on the homepage:

  • Smarter-than-human AI may come soon.
  • It could lead to human extinction.
  • Supporting arguments: AI is advancing fast, AI may pursue goals, AI can win a conflict against us, and experts are highly concerned.

If you ask Stampy about these topics, expect answers that reflect this framing. The homepage also links to "I need more arguments" for readers who want the reasoning laid out in more depth.

Limits and caveats

  • Accuracy: The site states plainly that Stampy can be inaccurate. Do not rely on it for precise claims without checking sources.
  • Scope: It is described as a tool for questions about AI safety, not a general-purpose assistant.
  • Verification: The instruction to "always verify its sources" implies answers may include references you are expected to check yourself.

Who runs the site

AISafety.info is a project founded by Rob Miles, maintained by a global team of specialists and volunteers. It is an Ashgro Inc project; Ashgro Inc (EIN: 88-4232889) is a 501(c)(3) public charity incorporated in Delaware. The site lists partner projects including AISafety.com and Alignment Ecosystem Development, and offers ways to get involved through code, writing, Discord, and donations.

Practical takeaway

Use Stampy when you want a quick orientation to an AI safety question and are willing to verify what it tells you. Use the site's articles and categories when you need something you can check and cite. If you want to contribute rather than just read, the "Get involved" and "How Can I Help?" paths are the relevant entry points.