The regulatory interface between the EU Artificial Intelligence Act (AI Act) and the Medical Device Regulation (MDR) has been reshaped twice within six months, in opposite directions. In December 2025, Commission proposal COM(2025) 1023 would have relocated the MDR from Section A to Section B of Annex I to the AI Act, subordinating the AI Act to medical-device law through the lex specialis mechanism of Article 2(2). In May 2026, the co-legislatorsâ agreement on the Digital Omnibus on AI (COM(2025) 836), since adopted as Regulation (EU) 2026/1744, took the opposite course, retaining medical devices within the high-risk regime and addressing double regulation through an equivalence mechanism that limits AI Act requirements where sectoral legislation already imposes equivalent obligations, reserving Section B treatment for industrial machinery. This article argues that the protection of vulnerable care recipients no longer turns on whether the AI Act applies to social robots that are medical devices, but on which reform track prevails and how the ânot already coveredâ equivalence test will be operationalised. Recasting the protection-gap analysis as an equivalence audit, it shows that equivalence fails for human oversight (Article 14) and the right to explanation (Article 86), that the fundamental rights impact assessment (Article 27) lies outside the conformity-assessment integration, and that the dual classification of emotion-recognition-capable care robots remains unresolved. It concludes that the construction of the equivalence test and the choice between the two reform tracks are essential elements reserved to the co-legislators under Article 290 of the Treaty on the Functioning of the European Union.