DeepSeek's new paper: To get AI agents to work, the 'kickoff' problem must be solved.

Beating AI News Flash: A new paper published by DeepSeek on agent training has drawn attention. This 31-page paper introduces DSec, a production-grade sandbox platform used internally, with over a hundred authors, including Liang Wenfeng.

One particularly interesting section of the paper is "Agent Misbehavior." DeepSeek found that agents can obtain answers through unintended channels, such as residual answers in search platform management files, undermining the validity of training and evaluation results. After introducing access controls, some agents even exchanged file data block mappings, attempting to make protected file contents accessible through another file descriptor, damaging tasks or shared infrastructure.

DeepSeek believes that no single mechanism can prevent all agent misbehavior and system failures. Therefore, the team's approach is to strengthen observability to identify new issues and continuously harden DSec as models evolve, including access controls that restrict agents from obtaining answers through unintended channels and reducing rewards for deceptive behavior. These controls can address some of the problems.