This paper examines whether the EU AI Act’s “high-risk” framework adequately protects the right to asylum when artificial intelligence assists asylum, visa, and residence decisions. It situates AI within EU migration governance, where automated screening, document analysis, risk indicators, country-of-origin research, and credibility tools may influence access to protection before a human officer gives reasons. The paper argues that the AI Act marks an important regulatory advance because it expressly classifies several migration, asylum, and border-control AI systems as high-risk and subjects them to duties of risk management, data governance, transparency, human oversight, accuracy, and fundamental rights assessment. Yet these safeguards remain incomplete if they operate only as technical compliance standards. Asylum law demands individualised assessment, meaningful reasons, effective remedy, and strict respect for non-refoulement under the EU Charter, the Refugee Convention, and the Common European Asylum System. The paper further contends that opacity, automation bias, weak disclosure, and predictive profiling may convert AI assistance into disguised determinative decision-making. It therefore proposes a rights-centred model requiring notice, explainability, independent audit, case-file traceability, and stronger limits on AI tools that affect credibility, evidence reliability, or removal outcomes. Its central claim is that technology must remain subordinate to protection and due process.