Bilingual language activation is generally non-selective, allowing both languages to be active simultaneously and creating the potential for cross-language interference. The present study examined whether the language background of a conversation partner influences the degree of cross-language interference experienced by German-French bilingual speakers. Scaling up a picture-word interference task to a social interaction between participants and confederate conversation partners, this study tested whether interacting with a bilingual partner, compared to a monolingual partner, increases L2 (French) interference and slows L1 (German) naming latencies. Preliminary results (n = 8) revealed an overall semantic interference effect, with slower responses for semantically related compared to unrelated trials. Consistent with our predictions, this semantic interference effect was larger with a bilingual than monolingual partner, as demonstrated by a promising, yet preliminary, marginally significant interaction effect. This modulation is what we refer to as partner-adapted language balance. Building on previous findings, we further examined whether participants with higher L2 proficiency would exhibit a more sensitive partner-adapted language balance. We found no significant effect of participants’ L2 proficiency, although the data showed a slight numerical trend toward larger semantic interference effects for participants with higher L2 proficiency, consistent with our predicted direction. Taken together, our preliminary findings suggest that a partner’s language background modulates cross-language activation, supporting the view that bilinguals flexibly adjust their situational language balance to match the current social context. These are interim results from the first eight participants; data collection will continue until the predetermined sample size (N=40) is reached, to allow a robust evaluation of the effects.