hacking-agents

Tracebit introduces 'context bombing,' a technique that uses prompt injections as defensive decoys to halt AI hacking agents, reducing admin compromise from 57% to 5% across tests with leading LLMs.