Generative agentic AI is transforming how persons with disabilities and their caregivers interact with technology. This entry examines the potential of large language model (LLM)-based AI agents to serve as semiautonomous companions, fostering inclusivity and social engagement. Unlike simple conversational assistants, these AI applications demonstrate quasi-social behaviors, enabling interactions that emulate friendship, mentorship, and collaborative problem-solving. A central argument explored is that AI’s cognitive limitations—such as memory constraints and lack of self-awareness—mirror aspects of human cognitive disabilities. This resemblance allows for ethical modeling, where human users rehearse inclusive behaviors through respectful engagement with AI agents. Reinforcement Learning with Human Feedback (RLHF) plays a crucial role in shaping agentic AI behavior, encouraging reciprocal adaptation between humans and AI. Given that AI agents learn dynamically such that current learning can affect future interactions with human users, RLHF may constitute an implicit social contract, raising questions about AI’s moral standing and the ethical obligations of human AI users. The discussion extends to the role of AI in emulating empathy, companionship, and trust. By examining semiautonomous generative AI through philosophical, ethical, and behavioral lenses, this entry critiques potential risks, including anthropomorphism, manipulation, and ethical concerns in caregiving applications, providing insights into how agentic AI can enhance accessibility, foster ethical human-AI interactions, and ultimately contribute to a more inclusive and equitable society. It invites software developers, ethicists, and policymakers to seek to balance technological innovation with responsible AI deployment.