individual
Technical AI safety researchAI safety meta
| BlueDot Impact | $3,500 | 100% | 1 |
| Date ↓ | Funder | Via | Amount | Cause | Source | Purpose |
|---|---|---|---|---|---|---|
| Jun 2026 | BlueDot Impact | $3,500 | Technical AI safety research, Epistemic infrastructure | bluedot | Testing whether subliminal learning in model distillation is driven by shared unembedding/readout alignment, and whether this risk appears in LLMs. |