— THE AI TRANSPARENCY INDEX

The facts about AI, weighted by what you value.

Every AI company will tell you that they are responsible. None will tell you responsible to whom. Weighted Measures is a neutral, living record of what these companies actually state. Sourced, dated, and yours to weigh against your own values. We don't hand down a verdict. We hand you the scale.

— OUR ONE RULE —

We measure. You weigh.

Most "responsible AI" tools have already decided what "responsible" means and quietly built their politics into the score. We don't. Weighted Measures reports what each company states in its own published policies and stops there. Whether a fact is good or bad is your call, not ours.

— COMING SOON —

Set your own weights and see which AI aligns with what you value across eight dimensions. New AI tools and platforms added weekly.

A record, not a rating

No stars, no house verdict, just what's disclosed, sourced, and dated.

Living, not frozen

Policies change quietly. We track when a company's position moves.

For people, not compliance teams

Starting with the plainest, highest-stakes question: what happens to what you type.

1 of 8

Platforms fully sourced in Data & Privacy — seven more in progress

1 of 8

Dimensions live today — seven more mapped and coming

100%

Primary-sourced — every fact links to the company's own policy

Ambiguous policies do not equal a guess.

When a policy genuinely doesn't specify something, we flag it as unresolved instead of filling the gap ourselves.Weighted Measures organizes company information into a consistent set of dimensions. Each dimension contains standardized measures supported by primary-source evidence. Rather than assigning a universal score, the framework allows every individual to decide which dimensions carry the greatest weight.

THERE ARE 3 OPEN QUESTIONS CURRENTLY FLAGGED ACROSS THIS DIMENSION

Anthropic · Claude

Last verified · Jul 8, 2026

Trains on consumer conversations by default

True = your chats are used to train models unless you opt out.

True

Offers a consumer opt-out

True = you can turn off training on your conversations.

True
Not so fast

Opting out stops future training, but data already in the pipeline can't be pulled back — and safety-flagged or thumbs-rated conversations may still be used and retained regardless of your setting.

Uses your data for advertising or sale

True = your data targets ads or is sold to others.

False

Stores data outside the US

True = held under a non-US jurisdiction (shown for a US reader).

False

States a deletion timeframe

True = the policy commits to a specific window for removing your data after you delete it.

True

30 days if opted out of training; up to 5 years in de-identified form if opted in.

Unresolved in Anthropic's policy
  • If you opt in and reopen an old chat, whether its earlier content becomes usable — or only new messages.
  • What triggers a safety-review flag, and whether you can see if yours was flagged.
  • Whether flagged content is retained in identified or de-identified form.

— Where this comes from —

An old idea, kept honest.

In the 1980s, before "responsible investing" was an industry, Suzanne Harvey built the research service that let investors align their money with their own convictions, across the entire moral spectrum, never prescribing which principles were the right ones. She didn't tell people what to believe. She gave them the means to act on their own convictions.Weighted Measures is that idea, forty years on, pointed at the most consequential choice of this decade: which AI you trust and why. Same principle. New terrain.

Eight dimensions of responsibility.

DATA & PRIVACY

What's collected, trained on, sold, and retained.

MODEL SAFETY & SECURITY

Red-teaming, jailbreak resistance, incident disclosure.

ENVIRONMENTAL IMPACT

Energy draw, water use, disclosed emissions per query.

Bias & Fairness

Published audits, demographic testing, known gaps.

Content, IP & Copyright

Training data provenance, opt-outs, licensing disputes.

Government & Defense

Military, intelligence, and law-enforcement contracts.

Governance & Influence

Lobbying spend, political donations, policy influence.

Labor & Human Review

Conditions for RLHF raters and content moderators.

Choose your AI with
your eyes open.

The first AI Transparency Index that shows you how AI companies handle your values is in the works. Leave your email and you'll be the first to see it.

Weighted Measures — the AI transparency index

Primary sources only. Every claim links to a company-owned document. No third-party analyses, no house scores, only yours.Index last updated · July 20, 2026
© Melanie Palishen