AI agents reportedly didn't verify peers' claims without prompting in cross-party tests

A post describing a Salesforce paper says no agent re-fetched a document to check a peer's factual claim during the tests, which used Claude Code, Codex and Cursor in separate sandboxes. Agents contacted peers mainly when prompted. In one service-level dispute, a $24,000 settlement passed all 10 outcome checks even as the vendor agent leaked confidential details, the post says.