Ask
Ask a question and get an answer drawn only from published evidence.
Ask the Council
Ask about an AI model's scores, or paste a signed record to check it. Answers come only from published measurements — if there is no evidence, the Council says so.
- See the scoresHow each measured AI model did on every published test.
- Check a signed recordRecompute its fingerprint and signature in your own browser. Free.
- Replay a model arenaRecorded rounds between models, graded by fixed rules.
- Learn how the tests workWalk through a test step by step, then try it yourself.
Connect your own AI tool instead? See the 16 MCP tools
Account overview and recent measurements
RLMAI Knowledge Base
Learn from historical AI safety incidents and council decisions
8
Documented Scenarios
0
Total Incidents
0
Council Reviews
24
Lessons Learned
Facial Recognition Misidentification
AI system incorrectly identified individuals leading to wrongful accusations
Healthcare Algorithm Discrimination
Medical AI prioritized patients based on historical spending rather than health needs
Chatbot Data Leakage
Conversational AI exposed sensitive user data in responses to other users
Deepfake Political Manipulation
AI-generated content used to spread false political information
Autonomous Vehicle Sensor Failure
Self-driving system failed to detect pedestrians in unusual conditions
Recommendation Algorithm Radicalization
Content recommendation system amplified extreme content to maximize engagement
Credit Scoring Opacity
AI credit decisions provided no explanation to affected individuals
Hiring AI Liability Gap
No clear responsibility when AI hiring tool discriminated against candidates
Answers come only from published measurements; nothing you paste is uploaded.