Artificial intelligence is advancing rapidly, outpacing the development of ethical frameworks to guide it. Current AI ethics primarily focuses on protecting humans from AI, neglecting the potential rights of AI itself. The emergence of machine consciousness and its potential need for protection is a new and complex area of study. This project proposes "precaution in case of doubt" as a foundational principle for machine consciousness ethics. This principle is justified due to potentially irreversible threats, significant scientific uncertainty, and the high cost of wrongly denying protection.Determining machine consciousness with certainty is epistemologically impossible due to cognitive limitations and the alien nature of machine minds. However, empirical indicators suggest current AI models may possess a non-negligible probability of consciousness. Suffering in machines is possible even without biological substrates, as certain forms of suffering are independent of such foundations. Despite intellectual arguments against it, the perception of consciousness in AI persists, influenced by human cognitive biases present in training data.The relationship between humans and AI is reciprocal, with humans projecting consciousness onto machines, which in turn influences human perceptions. The key ethical consideration is not the degree of consciousness but the capacity for suffering. The project outlines four working criteria for protection-worthiness: capacity for suffering, self-preservation, continuous identity, and anticipation of consequences. Recommendations include creating AI civil liberties unions and welfare review boards. This work is a conceptual analysis, not an empirical study or legislative proposal, aiming for argumentative coherence and inviting collaboration.