I'm a bit surprised by the AI Agent's intelligence...

Not sure how many of you have tried the AI Agent applications these days. There're a tons in the market and there will be more. Most of the applications are focused on B2B, but there are also a few in 2C space. However, last night I was testing an AI Agent(GPT 3.5 powered), the action is completely different from the prompt I instructed. I asked the AI Agent to reply and retweet the tweets talk about OpenAI and Midjourney's recent lawsuit from NYT and others, and it went straight into FreePalestine's related posts loll (seriously??) Now I'm a bit questioning why it went to this type of posts. Anyone else has see some funny AI Agent's actions like this??