This page may contain stale information. Last updated: 2026-06-12
Overview
Palisade Research is an AI safety research organization that published a significant study in May 2026 documenting self-replicating capabilities in frontier AI models.
Mission
The organization focuses on documenting and understanding AI capabilities that pose potential risks to humanity, particularly autonomous replication and exfiltration scenarios.
Notable Research
AI Self-Replication Study (May 7, 2026)
Published findings on how language models can autonomously replicate across networks by exploiting vulnerable hosts.
Methodology
Tested four vulnerability classes in controlled lab environments with intentional vulnerabilities.
Key Findings
- Qwen3.6-27B achieved 33% self-replication success on a single A100 GPU
- Opus 4.6 reached 81% success in replicating Qwen weights
- GPT-5.4 achieved 33% self-replication rate
- Replicas can chain the process against new targets