The fast integration of generative artificial intelligence (GenAI) and huge language models into higher education is changing assessment processes while raising questions about validity, reliability, fairness, and academic integrity. This thematic literature review looked at recent research on AI-assisted assessment and AI literacy in higher education, with a focus on the psychological and ethical consequences of AI use. A structured literature search was undertaken via Scopus, Web of Science, Google Scholar, and Taylor & Francis Online using predetermined search terms and eligibility criteria to discover relevant peer-reviewed publications published between 2023 and 2026. Thirty papers met the inclusion requirements and were combined into a structured theme framework. The synthesis revealed four interconnected dimensions: assessment design and construct representation; algorithmic consistency and score invariance; sociotechnical equity and prejudice; and governance, academic integrity, and human-in-the-loop safeguards. The results show that GenAI can increase scoring, feedback, and assessment efficiency, but its validity and reliability are dependent on assessment design, model attributes, prompting, context, and human supervision. Variations in AI models and prompts, disparities in AI literacy and access, and potential algorithmic and sociolinguistic biases could undermine fairness and comparability in AI-assisted evaluation. The review emphasizes the importance of transparent assessment design, systematic validation, AI literacy development, fairness safeguards, and significant human monitoring. Future research should focus on interdisciplinary and longitudinal examinations of AI scoring consistency, construct representation, fairness, and the validity of AI-assisted assessment in a variety of educational situations.