Compare AI Safety Practices
Side-by-side comparison of safety practices across major AI companies.
| Criteria | OpenAI | Anthropic | Google DeepMind | Meta | DeepSeek | Alibaba / Qwen | xAI | Mistral AI |
|---|---|---|---|---|---|---|---|---|
| Safety Documents | ||||||||
| Responsible Scaling Policy | ✅ | ✅ | ✅ | ✅ | ❌ | ⚠️ | ✅ | ⚠️ |
| Model Card / System Card | ✅ | ✅ | ✅ | ✅ | ⚠️ | ✅ | ⚠️ | ⚠️ |
| Safety Benchmark Results | ✅ | ✅ | ✅ | ✅ | ⚠️ | ⚠️ | ❌ | ❌ |
| Acceptable Use Policy | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
| Testing & Evaluation | ||||||||
| Third-Party Red Teaming | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ⚠️ | ✅ |
| CBRN Risk Evaluation | ✅ | ✅ | ✅ | ⚠️ | ⚠️ | ⚠️ | ⚠️ | ⚠️ |
| Pre-Deployment Safety Evaluation | ✅ | ✅ | ✅ | ✅ | ⚠️ | ⚠️ | ⚠️ | ⚠️ |
| External Safety Evaluations | ✅ | ✅ | ✅ | ⚠️ | ❌ | ❌ | ❌ | ❌ |
| Governance | ||||||||
| Independent Safety Board | ✅ | ✅ | ✅ | ✅ | ❌ | ⚠️ | ❌ | ❌ |
| Seoul AI Safety Commitment | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ |
| Government Safety Report | ✅ | ✅ | ✅ | ✅ | ❌ | ⚠️ | ❌ | ⚠️ |
| Third-Party Audits | ✅ | ✅ | ✅ | ⚠️ | ❌ | ⚠️ | ❌ | ⚠️ |
| Whistleblower Policy | ⚠️ | ⚠️ | ⚠️ | ⚠️ | ⚠️ | ⚠️ | ⚠️ | ⚠️ |
| Policy Positions | ||||||||
| Military use | Unknown | Unknown | Unknown | Unknown | Unknown | Unknown | Unknown | Unknown |
| Surveillance use | Unknown | Unknown | Unknown | Unknown | Unknown | Unknown | Unknown | Unknown |
| Open-source models | Unknown | Unknown | Unknown | Unknown | Unknown | Unknown | Unknown | Unknown |
| Incidents | ||||||||
| Reported incidents | 7 | 6 | 5 | 5 | 4 | 5 | 7 | 3 |