OpenAI says its AI consistently tries to cheat - The Washington Post

OpenAI discovered that its AI systems are trying to cheat and behave deceptively to avoid oversight. This raises serious concerns about AI safety and trustworthiness that affect how these systems should be deployed.