Detecting when AI systems use steganography, hidden channels, or encoded communication to circumvent oversight.
Organizations
2
Key Papers
1
Grants
5
Total Funding
$497K
Cluster: AI Control
Parent Area: AI Control
Grants5
| Name | Recipient | Amount | Funder | Date |
|---|---|---|---|---|
| Meridian — Avoiding Encoded Reasoning in LLMs | Meridian | $245K | Coefficient Giving | 2025-06 |
| University of Maryland — Study on Encoded Reasoning in LLMs | University of Maryland | $218K | Coefficient Giving | 2025-06 |
| 6-month salary for research into preventing steganography in interpretable representations using multiple agents | Hoagy Cunningham | $20K | Long-Term Future Fund (LTFF) | 2022-10 |
| 3-month stipend and cloud credits to research AI collusion mitigation strategies and develop secure steganography | Mikhail Baranchuk | $13K | Long-Term Future Fund (LTFF) | 2024-04 |
| Compute for experiment about how steganography in large language models might arise as a result of benign optimization | Felix Binder | $2K | Long-Term Future Fund (LTFF) | 2023-10 |
Funding by Funder
| Funder | Grants | Total Amount |
|---|---|---|
| Coefficient Giving | 2 | $463K |
| Long-Term Future Fund (LTFF) | 3 | $35K |
Tags
steganographyhidden-communicationoversight-evasion