Style Guide

Thanks for taking the time to help keep the scorecard accurate. Every edit you submit is reviewed by ASVA before it goes live; nothing you write is published automatically; following this guide makes it far more likely your suggestion is accepted as-is, and quickly.

1. Source every claim

The scorecard's credibility rests on every claim being checkable. For each bullet point you add or change, include a source; this could be a link to the MP's own statement, a vote record, a written question, a campaign signatory list, or comparable public record. A bullet with no source is the single most common reason a submission is sent back, so this alone will save you a round trip.

2. Keep it factual

3. Follow the format

4. Grading rubric

ASVA tracks engagement with catastrophic and extinction-level AI risk and superintelligence specifically, not AI policy in general. Without a position on, or work on, superintelligence or extinction-level risk, no MP can be graded above a C, however active they are on AI more broadly.

The things that count as an x-risk signal are:

Use this table to see which grade the evidence supports:

GradeWhat it looks like
ASigned the Sobel letter and either ControlAI or the PauseAI letter to the PM; or one x-risk signal plus active work on x-risk specifically, e.g. co-sponsoring the AI kill-switch amendment, leading a debate on extinction-level risk, or sponsoring a superintelligence bill. General AI activity (copyright, jobs, the NHS, deepfakes) does not lift a card to A.
BOne x-risk signal (see the list above) plus whatever general AI record they have, e.g. a ControlAI signatory with little else on record, or an MP who posted backing a superintelligence ban.
CNo x-risk signal, but has pushed for regulation of AI in some way, e.g. called for an AI regulator or statutory safeguards, backed a bill or amendment regulating AI (even a narrow one, such as on AI chatbots and children), or pressed ministers to bring forward AI legislation. Warning that AI is dangerous is not enough on its own; the MP must have asked for AI to be regulated, overseen or limited.
?Not scored. There isn't enough public information, or the MP's AI record is neutral or mixed: questions about using or adopting AI, AI and copyright on its own, or warnings about AI risks without any push for regulation. This is not a negative grade.
DPuts growth or innovation ahead of safeguards without campaigning against them, e.g. championing the UK as an "AI powerhouse", deferring promised safeguards, or rejecting a superintelligence ban as stifling innovation.
EActively works against AI safety governance, e.g. urging cuts to AI regulation, attacking AI rules as red tape, or calling a pause on AI a betrayal.
FClear, vocal dismissal of AI risk and its regulation, e.g. calling AI risk "hype" or "doomerism", or saying a slowdown or pause is impossible because AI is a race to win.
DNRDid not respond. Set by ASVA only, after it has contacted the MP and had no reply; please don't propose it.

A low grade needs evidence just as a high one does: never infer opposition from silence. If you're unsure between two grades, propose the lower one and explain why in your bullets; the reviewer makes the final call.

5. What not to submit

What happens after you submit

An ASVA reviewer checks your suggested edit against the sources you provided. You'll get an email either way: if it's published, whether it went live exactly as you wrote it or with some changes; if it isn't, a short note on why not so you can try again.