Compassion Benchmark
Insights

Insights

Argued, cited pieces on compassion-first AI for AI labs, agent builders, press and policy readers. Every figure carries a source or is derived from the data published on this site.

Posts

Agent builders · 7 min read

A compassion checklist for people building AI agents

Agents act, so their failures are actions as well as words. A practical checklist that maps the benchmark's compassion dimensions to concrete agent design decisions.

Labs and agent builders · 7 min read

How to read an AI model release for welfare and safety signals

A system card or release note can be read for specific, checkable disclosures. Here is what to look for, what is often missing, and what to ask.

AI labs · 7 min read

The case for compassion-first AI, with what we know and what we don't

The credible case rests on duty of care, regulation, liability and procurement. The commercial upside is plausible but unproven, and we say where the evidence runs out.

Labs and agent builders · 6 min read

What "compassion-first AI" means, and what it does not

Compassion-first AI is a set of observable behaviours for recognising, responding to and reducing suffering, not a warm tone, agreeableness or a refusal policy.

AI labs · 6 min read

Why AI labs should be assessed on organisational compassion, not only model behaviour

A model is one snapshot of what a lab ships; governance, incident response and disclosure are decided by the organisation, and they need their own measurement.

Labs and agent builders · 6 min read

Why we do not rank AI models yet, and what would change that

Declining to rank models is a methodological position with stated conditions, set by a public item bank, a single judge family and no human validation yet.

Compassion Benchmark is independent. No provider pays for inclusion, coverage, status or wording. Commercial services support access and interpretation only, never findings.

See an error? Send a correction. Corrections are published as new dated records and the original stays visible.

Email me when a new piece publishes

One email when a new Insights piece or AI Model Watch edition is published. No promise of weekly timing.

Free. No promise of weekly timing. We never share your email.