Anthropic is a $380B-valued AI company with virtually no environmental accountability. It reports zero emissions data despite being one of the world's largest compute consumers, projects 5GW data center needs by 2028, and advocates for weaker environmental permitting rules. No targets, no renewable energy strategy, no transparency.
Same formula for every company. No curve. No private weighting.
SINK = (0.3 × Base + 0.7 × Performance) × ScaleStrongest on Controversies & Red Flags and Resource Use & Waste (7/10, 2/10). Weakest on Emissions Trajectory and Transparency & Accountability (0/10, 1/10).
8 sources used in this assessment. All publicly available. Each row shows which rubric questions it informed.
If you believe a source has been misread or a newer version exists, submit a challenge.
Among the 35 major saas / digital services brands we've scored, Anthropic sits 34th of 35.
Score history begins 6 April 2026.
As Anthropic's score updates, the trajectory will appear here.
We're backfilling historical scores for FTSE 100 and S&P 100 companies over the coming weeks.
Every challenge is published. We'd rather be corrected than wrong — that's the whole point.
No challenges submitted yet. If you have evidence that contradicts this score, you can challenge any question above — cite a public source and we'll review it.
Anthropic is a San Francisco-based AI safety company founded in 2021, developing large language models including Claude. With ~3,000 employees and rapidly growing revenue, it is one of the largest consumers of cloud compute infrastructure globally, relying primarily on AWS and Google Cloud for its computational needs.
Comparable AI giant with similar scale, compute intensity, and environmental non-disclosure patterns.
View breakdown →Major Anthropic funder and cloud provider; faces scrutiny over AI data center energy and water impacts.
View breakdown →AWS supplier to Anthropic; large data center operator facing water and emissions accountability demands.
View breakdown →Email alerts when a rubric question is verified, a challenge is resolved, or the overall score changes.
One email, every Sunday. Score changes, new research, the stories behind the numbers. Free.
No spam. Unsubscribe in one click.
Readers and institutions support our work. Companies can pay to submit evidence we couldn't find. Neither type of payment changes a score.