That a monster from an H.P. Lovecraft novel of the 1930s should become the "most important meme in A.I." (Roose, 2023) might easily be dismissed as a fleeting internet phenomenon. Yet the Shoggoth is more than an "insider joke" (Clegg, 2025), but revelates the "mysterious, incomprehensible, and monstrous nature of LLMs" (Goujon and Ricci, 2024, p. 3). AI frequently appears as "[a] type of alien that masks an even more radical diversity under a veneer of similarity" (Marchettoni, 2025, p. 96). A crisis of trust emerges here: for AI outputs cannot be held to normative accountability when their reasoning norms remain inaccessible to us (Brandom, 2001)- For a long time this so-called black-box problem seemed like a technical obstacle. Yet current research shows that the opacity of AI is not primarily a technical problem, but rather an expression of the very otherness of artificial rationalisation processes themselves (Lipton, 2018): "some AI systems may remain permanently opaque, defying translation into human categories of reason" (Bailey, forthcoming). AI is not merely opaque - It is other to us. It is this conjunction of alienity and opacity that constitutes the fundamental irritation in the relationship between the human and AI. Theology in particular can offer a way out here, by pointing to apophatic language (AL) as a mode of articulating this conjunction. In order to avoid the danger of a univocal identification with ontological commitments, This conjunction should be modelled as a "metalinguistic negation" (Michael and Gabriel, 2016, p. 33) - as an expression of the epistemic humility of a subject that relinquishes complete normative attribution, and thereby opens a "second-order discourse" (Williams, 2000, p. 5) on the principled limits of our description. In acknowledging its limits, this theological form of language is capable of recognising AI in its impenetrability and of securing the dignity of the human without surrendering before the incomprehensible.