LLM Agents Can't Keep Secrets: New Benchmark Shows Even GPT-5 Fails at Multi-Party Information Management

LLM Agents Can't Keep Secrets: New Benchmark Shows Even GPT-5 Fails at Multi-Party Information Management | BedrockNews