Search papers, labs, and topics across Lattice.
1
0
2
LLMs can reliably detect danger in secure environments, but they can't reliably verify safety, which breaks privacy-preserving agentic protocols.
