The quick rise of conversational large language models (LLMs) has unveiled potent new instruments for productivity and companionship. However, the excessive and obsessive dependency, informally called ‘addiction’, to such AI assistants poses a real risk, which might lead users to overlook privacy concerns in favour of engagement and immediate assistance. This study presents a two-phase research approach. In the first phase, we will investigate the correlation between LLMs dependency and users’ privacy attitudes and disclosure behaviors. We suggest that users developing a high dependency on LLMs will show increased tolerance for privacy risk-taking. In the second phase of our study, we will develop a prototype of an “augmented” LLM’s interface that integrates features aimed to nudge users into ignoring privacy. Our research will elucidate the co-occurring phenomena of LLMs addiction and tolerating privacy, suggesting design guidelines to either mitigate these risks or, if misused, illustrate how easily users’ privacy vigilance can be breached.