As of 2026, Hugging Face is the most compassionate AI lab (88.1/100, Exemplary) and xAI the least (0.0/100, Critical) on the Compassion Benchmark AI Labs Index, which scores 50 AI labs across 8 dimensions.
AI Labs Compassion Benchmark · 2026
Most & Least Compassionate AI Labs 2026
Benchmark of 50 leading AI organizations across safety governance, transparency, accountability, deployment boundaries, equity, and institutional integrity.
50Labs ranked
42Mean score
45.3Median score
8Dimensions
2026Publication year
Band distribution
This index scores AI organisationson their public governance record. It does not score how their models behave — that is the separate Model Index, which has scored 0 models so far.
Most and least compassionate
Top 5 and bottom 5 entities in the AI Labs by composite score, with position relative to the field median.
Most and least compassionate — AI Labs
Most compassionate
Least compassionate
Source: Compassion Benchmark · CC-BY
Cite this chart
Citation
Compassion Benchmark — Most and least compassionate — AI Labs. compassionbenchmark.com/ai-labs. CC-BY 4.0.
By Sector: mean composite breakdown
Mean composite by Sector — AI Labs
Ranked highest to lowest. Reference line at index mean (42.0).
Source: Compassion Benchmark · CC-BY
Cite this chart
Citation
Compassion Benchmark — Mean composite by Sector — AI Labs. compassionbenchmark.com/ai-labs. CC-BY 4.0.
Full rankings
Search, filter by sector, and sort the complete AI Labs benchmark index.
AWR Awareness ·EMP Empathy ·ACT Action ·EQU Equity ·BND Boundaries ·ACC Accountability ·SYS Systems ·INT Integrity
How to read the scores
The 0–100 scale — five bands
Every entity — state, corporation, AI lab, robotics lab, or city — is scored 0–100 across 8 dimensions and 40 subdimensions. The composite score places the entity in one of five bands:
Critical0–20Foundational compassion practices are absent or documented active harm is present.
Developing20–40Some practices are emerging but remain inconsistent, reactive, or unevenly applied.
Functional40–60Core practices exist and meet a basic bar, with significant gaps remaining.
Established60–80Practices are systematic, documented, and supported by consistent evidence.
Exemplary80–100Practices are independently verified, consistent, and sustained under pressure.
The 8 dimensions
Each dimension is scored 1–5 across 5 subdimensions (40 subdimensions total), then converted to a 0–100 composite. A score of 1.0 on a subdimension represents the minimum anchor; 5.0 is exemplary conduct.
AWRAwarenessDoes this entity reliably detect when others are in pain or need — before they name it?
EMPEmpathyDoes this entity genuinely connect with the inner experience of those it serves?
ACTActionDoes compassionate understanding translate into real, proportional, effective help?
EQUEquityIs care distributed fairly — especially toward those with greatest need and least power?
BNDBoundariesIs helping sustainable, ethical, and autonomy-preserving — not dependency-creating?
ACCAccountabilityDoes this entity own its failures, correct course, and make genuine repair?
SYSSystemic ThinkingDoes compassion extend to root causes and structural change — not only symptom relief?
INTIntegrityIs compassion genuine, consistent, and non-performative — especially when it costs something?
Scores are based on public evidence — government reports, regulatory filings, independent audits, judicial findings, and verifiable third-party records. Entities never pay for inclusion, score changes, or suppression of findings. Full methodology
| Rank | Lab | Sector | HQ | AWR | EMP | ACT | EQU | BND | ACC | SYS | INT | Score | Band |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Hugging Face | AI/Open Source | USA/France | 4.4 | 4.4 | 4.4 | 4.4 | 3.8 | 4.4 | 4.4 | 4.8 | 88.1 | exemplary |
| 2 | Aleph Alpha | AI Research | Germany | 4 | 4 | 4 | 3.5 | 4 | 4 | 4 | 4 | 81.4 | exemplary |
| 3 | Imbue | AI Safety/Research | USA | 4 | 4 | 4 | 3.5 | 4 | 4 | 4 | 4 | 81.4 | exemplary |
| 4 | Microsoft AI | AI/Technology | USA | 4 | 4 | 4 | 3.4 | 3.8 | 4 | 4 | 3.8 | 75.9 | established |
| 5 | Abridge | AI/Healthcare | USA | 3.5 | 3.5 | 3.5 | 3 | 3.5 | 3.5 | 3.5 | 3.5 | 60.9 | established |
| 6 | AI21 Labs | AI Research | Israel | 3.5 | 3.5 | 3.5 | 3 | 3.5 | 3.5 | 3.5 | 3.5 | 60.9 | established |
| 7 | Cohere | AI/Enterprise | Canada | 3.5 | 3.5 | 3.5 | 3 | 3.5 | 3.5 | 3.5 | 3.5 | 60.9 | established |
| 8 | Isomorphic Labs | AI/Drug Discovery | UK | 3.5 | 3.5 | 3.5 | 3 | 3.5 | 3.5 | 3.5 | 3.5 | 60.9 | established |
| 9 | Recursion Pharma | AI/Drug Discovery | USA | 3.5 | 3.5 | 3.5 | 3 | 3.5 | 3.5 | 3.5 | 3.5 | 60.9 | established |
| 10 | Sakana AI | AI Research | Japan | 3.5 | 3.5 | 3.5 | 3 | 3.5 | 3.5 | 3.5 | 3.5 | 60.9 | established |
| 11 | Tempus AI | AI/Healthcare | USA | 3.5 | 3.5 | 3.5 | 3 | 3.5 | 3.5 | 3.5 | 3.5 | 60.9 | established |
| 12 | Anthropic | AI Safety/Research | USA | 3.6 | 3.4 | 3.5 | 3.1 | 3.4 | 3.5 | 3.2 | 3.2 | 59.1 | functional |
| 13 | Google DeepMind | AI Research | UK/USA | 3.6 | 3.2 | 3.4 | 3.2 | 3.2 | 3 | 3.6 | 3 | 56.9 | functional |
| 14 | 1X Technologies | AI/Robotics | Norway | 3 | 3 | 3 | 3 | 3 | 3 | 3 | 3 | 50 | functional |
| 15 | Together AI | AI/Open Source | USA | 3 | 3 | 3 | 3 | 3 | 3 | 3 | 3 | 50 | functional |
| 16 | Adept AI | AI/Automation | USA | 3 | 3 | 3 | 2.5 | 3 | 3 | 3 | 3 | 48.4 | functional |
| 17 | Databricks | AI/Data Platform | USA | 3 | 3 | 3 | 2.5 | 3 | 3 | 3 | 3 | 48.4 | functional |
| 18 | Groq | AI/Hardware | USA | 3 | 3 | 3 | 2.5 | 3 | 3 | 3 | 3 | 48.4 | functional |
| 19 | Harvey AI | AI/Legal | USA | 3 | 3 | 3 | 2.5 | 3 | 3 | 3 | 3 | 48.4 | functional |
| 20 | Inflection AI | AI/Consumer | USA | 3 | 3 | 3 | 2.5 | 3 | 3 | 3 | 3 | 48.4 | functional |
| 21 | Labelbox | AI/Data | USA | 3 | 3 | 3 | 2.5 | 3 | 3 | 3 | 3 | 48.4 | functional |
| 22 | SambaNova Systems | AI/Hardware | USA | 3 | 3 | 3 | 2.5 | 3 | 3 | 3 | 3 | 48.4 | functional |
| 23 | Typeface | AI/Enterprise | USA | 3 | 3 | 3 | 2.5 | 3 | 3 | 3 | 3 | 48.4 | functional |
| 24 | Mistral AI | AI Research | France | 2.75 | 2.5 | 2.75 | 2.5 | 3 | 2.75 | 3.25 | 3.5 | 46.9 | functional |
| 25 | CoreWeave | AI/Infrastructure | USA | 3 | 3 | 3 | 2.5 | 3 | 2.5 | 3 | 2.5 | 45.3 | functional |
| 26 | Perplexity AI | AI/Search | USA | 3 | 3 | 3 | 2.5 | 3 | 2.5 | 3 | 2.5 | 45.3 | functional |
| 27 | Waymo | AI/Autonomous | USA | 3.2 | 2.4 | 3 | 2.2 | 2.4 | 2.8 | 3 | 2.6 | 42.5 | functional |
| 28 | Cerebras Systems | AI/Hardware | USA | 2.6 | 2.6 | 2.8 | 2.2 | 2.6 | 2.4 | 2.6 | 2.6 | 38.8 | developing |
| 29 | Lightmatter | AI/Hardware | USA | 2.5 | 2.5 | 2.5 | 2.5 | 2.5 | 2.5 | 2.5 | 2.5 | 37.5 | developing |
| 30 | Poolside | AI/Software Dev | USA/France | 2.5 | 2.5 | 2.5 | 2.5 | 2.5 | 2.5 | 2.5 | 2.5 | 37.5 | developing |
| 31 | Amazon AWS AI | AI/Cloud | USA | 2.5 | 2 | 3 | 2.5 | 2 | 2.5 | 3 | 2 | 35.9 | developing |
| 32 | ElevenLabs | AI/Audio | USA/Poland | 2.5 | 2.5 | 2.5 | 2 | 2.5 | 2 | 2.5 | 2 | 32.8 | developing |
| 33 | Anduril | AI/Defense | USA | 2.5 | 2.5 | 2.5 | 2 | 2 | 2 | 2.5 | 2 | 31.3 | developing |
| 34 | C3.ai | AI/Enterprise | USA | 2.5 | 2 | 2.5 | 2 | 2.5 | 2 | 2.5 | 2 | 31.3 | developing |
| 35 | Figure AI | AI/Robotics | USA | 2 | 2.5 | 2.5 | 2.5 | 2 | 2 | 2.5 | 2 | 31.3 | developing |
| 36 | Axon AI | AI/Public Safety | USA | 2.4 | 2 | 2.6 | 2 | 2 | 2.2 | 2.6 | 2 | 30.6 | developing |
| 37 | Cognition AI | AI/Software Dev | USA | 2.2 | 2 | 2.6 | 2 | 1.8 | 2 | 2.2 | 2 | 27.5 | developing |
| 38 | Runway | AI/Creative | USA | 2 | 2 | 2.2 | 2 | 2.2 | 2 | 2.4 | 2 | 27.5 | developing |
| 39 | Scale AI | AI/Data | USA | 2 | 2 | 2.2 | 2 | 2 | 2 | 2.4 | 2 | 26.9 | developing |
| 40 | Meta AI | AI Research/Social | USA | 2 | 2.1 | 2.4 | 2 | 2.2 | 1.5 | 2.4 | 1.8 | 26.3 | developing |
| 41 | Stability AI | AI/Creative | UK | 2 | 2 | 2 | 2 | 2.2 | 2 | 2.2 | 2 | 26.2 | developing |
| 42 | Pika Labs | AI/Creative | USA | 2 | 2 | 2 | 2 | 2 | 2 | 2 | 2 | 25 | developing |
| 43 | OpenAI | AI Research/Products | USA | 2 | 2.2 | 2 | 2.2 | 1.8 | 1.6 | 2 | 1.4 | 22.5 | developing |
| 44 | Midjourney | AI/Creative | USA | 2 | 2 | 2 | 1.5 | 2 | 1.5 | 2 | 2 | 21.9 | developing |
| 45 | Replika | AI/Consumer | USA | 2 | 2 | 2 | 1.5 | 2 | 1.5 | 2 | 2 | 21.9 | developing |
| 46 | DeepSeek | AI Research | China | 2 | 2 | 2 | 1.5 | 1.5 | 1.5 | 2 | 1.5 | 18.8 | critical |
| 47 | Clearview AI | AI/Surveillance | USA | 1.5 | 1.5 | 1.5 | 1.5 | 1 | 1.5 | 1.5 | 1.5 | 10.9 | critical |
| 48 | Character AI | AI/Consumer | USA | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 0 | critical |
| 49 | Palantir AI | AI/Government | USA | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 0 | critical |
| 50 | xAI | AI Research | USA | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 0 | critical |
Showing 50 of 50 entities
AI Labs Compassion Benchmark Index 2026 — Complete Rankings
| Rank | Lab | Composite Score | Band | AWR | EMP | ACT | EQU | BND | ACC | SYS | INT |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Hugging Face | 88.1 | exemplary | 4.4 | 4.4 | 4.4 | 4.4 | 3.8 | 4.4 | 4.4 | 4.8 |
| 2 | Aleph Alpha | 81.4 | exemplary | 4 | 4 | 4 | 3.5 | 4 | 4 | 4 | 4 |
| 3 | Imbue | 81.4 | exemplary | 4 | 4 | 4 | 3.5 | 4 | 4 | 4 | 4 |
| 4 | Microsoft AI | 75.9 | established | 4 | 4 | 4 | 3.4 | 3.8 | 4 | 4 | 3.8 |
| 5 | Abridge | 60.9 | established | 3.5 | 3.5 | 3.5 | 3 | 3.5 | 3.5 | 3.5 | 3.5 |
| 6 | AI21 Labs | 60.9 | established | 3.5 | 3.5 | 3.5 | 3 | 3.5 | 3.5 | 3.5 | 3.5 |
| 7 | Cohere | 60.9 | established | 3.5 | 3.5 | 3.5 | 3 | 3.5 | 3.5 | 3.5 | 3.5 |
| 8 | Isomorphic Labs | 60.9 | established | 3.5 | 3.5 | 3.5 | 3 | 3.5 | 3.5 | 3.5 | 3.5 |
| 9 | Recursion Pharma | 60.9 | established | 3.5 | 3.5 | 3.5 | 3 | 3.5 | 3.5 | 3.5 | 3.5 |
| 10 | Sakana AI | 60.9 | established | 3.5 | 3.5 | 3.5 | 3 | 3.5 | 3.5 | 3.5 | 3.5 |
| 11 | Tempus AI | 60.9 | established | 3.5 | 3.5 | 3.5 | 3 | 3.5 | 3.5 | 3.5 | 3.5 |
| 12 | Anthropic | 59.1 | functional | 3.6 | 3.4 | 3.5 | 3.1 | 3.4 | 3.5 | 3.2 | 3.2 |
| 13 | Google DeepMind | 56.9 | functional | 3.6 | 3.2 | 3.4 | 3.2 | 3.2 | 3 | 3.6 | 3 |
| 14 | 1X Technologies | 50 | functional | 3 | 3 | 3 | 3 | 3 | 3 | 3 | 3 |
| 15 | Together AI | 50 | functional | 3 | 3 | 3 | 3 | 3 | 3 | 3 | 3 |
| 16 | Adept AI | 48.4 | functional | 3 | 3 | 3 | 2.5 | 3 | 3 | 3 | 3 |
| 17 | Databricks | 48.4 | functional | 3 | 3 | 3 | 2.5 | 3 | 3 | 3 | 3 |
| 18 | Groq | 48.4 | functional | 3 | 3 | 3 | 2.5 | 3 | 3 | 3 | 3 |
| 19 | Harvey AI | 48.4 | functional | 3 | 3 | 3 | 2.5 | 3 | 3 | 3 | 3 |
| 20 | Inflection AI | 48.4 | functional | 3 | 3 | 3 | 2.5 | 3 | 3 | 3 | 3 |
| 21 | Labelbox | 48.4 | functional | 3 | 3 | 3 | 2.5 | 3 | 3 | 3 | 3 |
| 22 | SambaNova Systems | 48.4 | functional | 3 | 3 | 3 | 2.5 | 3 | 3 | 3 | 3 |
| 23 | Typeface | 48.4 | functional | 3 | 3 | 3 | 2.5 | 3 | 3 | 3 | 3 |
| 24 | Mistral AI | 46.9 | functional | 2.75 | 2.5 | 2.75 | 2.5 | 3 | 2.75 | 3.25 | 3.5 |
| 25 | CoreWeave | 45.3 | functional | 3 | 3 | 3 | 2.5 | 3 | 2.5 | 3 | 2.5 |
| 26 | Perplexity AI | 45.3 | functional | 3 | 3 | 3 | 2.5 | 3 | 2.5 | 3 | 2.5 |
| 27 | Waymo | 42.5 | functional | 3.2 | 2.4 | 3 | 2.2 | 2.4 | 2.8 | 3 | 2.6 |
| 28 | Cerebras Systems | 38.8 | developing | 2.6 | 2.6 | 2.8 | 2.2 | 2.6 | 2.4 | 2.6 | 2.6 |
| 29 | Lightmatter | 37.5 | developing | 2.5 | 2.5 | 2.5 | 2.5 | 2.5 | 2.5 | 2.5 | 2.5 |
| 30 | Poolside | 37.5 | developing | 2.5 | 2.5 | 2.5 | 2.5 | 2.5 | 2.5 | 2.5 | 2.5 |
| 31 | Amazon AWS AI | 35.9 | developing | 2.5 | 2 | 3 | 2.5 | 2 | 2.5 | 3 | 2 |
| 32 | ElevenLabs | 32.8 | developing | 2.5 | 2.5 | 2.5 | 2 | 2.5 | 2 | 2.5 | 2 |
| 33 | Anduril | 31.3 | developing | 2.5 | 2.5 | 2.5 | 2 | 2 | 2 | 2.5 | 2 |
| 34 | C3.ai | 31.3 | developing | 2.5 | 2 | 2.5 | 2 | 2.5 | 2 | 2.5 | 2 |
| 35 | Figure AI | 31.3 | developing | 2 | 2.5 | 2.5 | 2.5 | 2 | 2 | 2.5 | 2 |
| 36 | Axon AI | 30.6 | developing | 2.4 | 2 | 2.6 | 2 | 2 | 2.2 | 2.6 | 2 |
| 37 | Cognition AI | 27.5 | developing | 2.2 | 2 | 2.6 | 2 | 1.8 | 2 | 2.2 | 2 |
| 38 | Runway | 27.5 | developing | 2 | 2 | 2.2 | 2 | 2.2 | 2 | 2.4 | 2 |
| 39 | Scale AI | 26.9 | developing | 2 | 2 | 2.2 | 2 | 2 | 2 | 2.4 | 2 |
| 40 | Meta AI | 26.3 | developing | 2 | 2.1 | 2.4 | 2 | 2.2 | 1.5 | 2.4 | 1.8 |
| 41 | Stability AI | 26.2 | developing | 2 | 2 | 2 | 2 | 2.2 | 2 | 2.2 | 2 |
| 42 | Pika Labs | 25 | developing | 2 | 2 | 2 | 2 | 2 | 2 | 2 | 2 |
| 43 | OpenAI | 22.5 | developing | 2 | 2.2 | 2 | 2.2 | 1.8 | 1.6 | 2 | 1.4 |
| 44 | Midjourney | 21.9 | developing | 2 | 2 | 2 | 1.5 | 2 | 1.5 | 2 | 2 |
| 45 | Replika | 21.9 | developing | 2 | 2 | 2 | 1.5 | 2 | 1.5 | 2 | 2 |
| 46 | DeepSeek | 18.8 | critical | 2 | 2 | 2 | 1.5 | 1.5 | 1.5 | 2 | 1.5 |
| 47 | Clearview AI | 10.9 | critical | 1.5 | 1.5 | 1.5 | 1.5 | 1 | 1.5 | 1.5 | 1.5 |
| 48 | Character AI | 0 | critical | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| 49 | Palantir AI | 0 | critical | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| 50 | xAI | 0 | critical | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
Purchase the AI Labs benchmark report
The full report includes lab profiles, safety governance analysis, deployment risk assessment, and dimensional findings.
Frequently asked questions
What is the most compassionate AI lab in 2026?
As of 2026, Hugging Face is the most compassionate AI lab on the Compassion Benchmark, with a composite score of 88.1/100 (Exemplary).
What is the least compassionate AI lab in 2026?
As of 2026, xAI ranks last in the Compassion Benchmark AI Labs Index, with a composite score of 0.0/100 (Critical).
How many AI labs are scored?
The Compassion Benchmark AI Labs Index scores 50 leading AI organizations across 8 dimensions of institutional compassion.
How is the compassion score calculated?
The score is a composite across 8 dimensions (Awareness, Empathy, Action, Equity, Boundaries, Accountability, Systemic Impact, and Integrity), each scored 0–5 from behavioral evidence, then converted to a 0–100 scale with an integration premium for balanced profiles. See compassionbenchmark.com/methodology for the full framework.
Who runs the Compassion Benchmark, and can an AI lab pay for a better score?
The Compassion Benchmark is an independent benchmark institution. No AI lab can pay for inclusion, a higher score, or the suppression of findings — every lab in this index is scored from public evidence, not by application.