OpenAI’s GPT-4 model is more trustworthy than GPT-3.5 but easier to trick – The Verge
Microsoft researchers found previously unpublished vulnerabilities around the trustworthiness of OpenAI's GPT models. Consumers, though, are not ...