This paper focuses on the ethical aspects of AIgenerated disinformation in Bangladesh, a new democracy and part of the Global South. This is a qualitative empirical study on how algorithmic content generation and dissemination, in this instance Facebook's AI algorithms, impact journalism and public debate. The study pairs 20 semistructured interviews with editors, journalists, technologists, and fact-checkers with a case study of 2020-2025 AIgenerated or manipulated content. The findings demonstrate that Facebook's algorithms promote sensational, biased and unethical content for engagement at the expense of truth and accuracy, and thus destroy ethical values like truth, responsibility and justice. The study also finds entrenched disinformation on a large number of topics with massive spikes closer to the 2024 elections. Activities of fact-checkers, media and regulators are mostly reactive and isolated from the larger scene with huge loopholes in accountability. By locating the issues of AI and media ethics within the context of a new democracy with weak institutions and polarized politics, the present research offers something to the worldwide debates. Theoretically, it uses Media Ethics Theory and Technological Determinism as a way to show how algorithms are degrading ethics. Practically, it proposes an integrated framework of ethics, law, and technology in Bangladesh media and provides new insights for democratic communication and the public trust.