The surge of generative AI (GenAI) tools on the internet has caused unprecedented disruptions to education, and there are growing concerns about them being misused by students to commit plagiarism, hence jeopardizing assessment validity. Despite the foreseeable challenges, it remains unclear how frontline language teachers perceive and react to the potential impact of GenAI on their assessment practice. To explore this, the current study interviewed 17 English language teachers teaching at 10 universities across China. Qualitative analyses of the data revealed that more than half of the participants perceived threats from GenAI to their assessment practice either now or in the future. Meanwhile, there was a lack of guidance and support from their institutions on how to deal with those challenges. Consequently, most of them had to come up with their own coping strategies, including employing detection methods, renovating existing assessments, and promoting AI ethics among students. However, doubts and concerns also arose over the weaknesses of some strategies, demonstrating teachers' struggles for optimal solutions. Implications for future assessment design are discussed.