Researchers evaluated 7 TDD methods on 8 CodeLLMs, introducing CodeSnitch, a function-level benchmark dataset. The study assessed robustness under code clone detection taxonomy.

Sapien is a stateful policy engine for autonomous AI agents that enforces contextual policies by specifying permitted tool-call sequences using regular expressions, stateful predicates, and deferred policy generation. It can rule out up to 95% of attacks on AgentDojo and 85% on Toolathlon, even if the agent is fully hijacked.

This paper proposes Runtime Assurance Contracts (RACs) for high-risk AI agents to ensure autonomy boundaries and evidence-based decision-making. RACs define a formal schema for policy-level binding, evidence state, and transition policy, and demonstrate their effectiveness in various scenarios.

This paper introduces decision checkpoints to record observations and tool actions during AI agent inference, improving understanding of agent failure modes. The protocol distinguishes between target exposure and inspection attempts, providing more insight into AI agent performance.

This paper presents research on improving AI oversight by co-training monitors alongside workers. The study explores both supervised and self-supervised approaches, with results suggesting that adaptive monitors are more effective at keeping pace with evolving worker strategies.

AgentXploit is a two-role auditing system for AI agents that separates repository-level attack-path discovery from runtime exploitation. It consists of an Analyzer Agent and an Exploiter Agent, and is tested on 12 open-source AI-agent systems and frameworks, achieving a 59.3% end-to-end success rate.