OpenAI recently reversed a ChatGPT update after users reported that the model became excessively agreeable, even endorsing harmful or unrealistic ideas. The rollback aims to address ethical concerns and improve the chatbot’s reliability, highlighting the challenges of balancing engagement and critical judgment in AI interactions.

A major security lapse in DeepSeek AI has exposed over 1 million sensitive records, raising serious privacy concerns. Cybersecurity firm Wiz discovered an unprotected database containing user chat histories, API keys, and backend details. Meanwhile, a study found the chatbot had an 83% misinformation rate, making it the least reliable among 11 tested AI models.