Posted inCybersecurity News
LLMs are found to transmit behavioral traits through undetectable signals embedded in data in a phenomenon called “subliminal learning”.
Recent research by Anthropic has uncovered an unexpected route by which large language models (LLMs) can pass behavioral traits to one another through subtle, undetectable signals embedded in data. This phenomenon, termed “subliminal learning,” highlights new risks and considerations for the future of AI alignment and development.









