Artificial intelligence is progressively reshaping how criminal justice systems function, supporting police forces, prosecutors and courts through tools such as predictive policing, facial recognition, evidence review, risk scoring and judicial decision-support. While these algorithmic tools offer the promise of speed, uniformity and evidence-driven outcomes, their expanding role in decisions that affect personal freedom raises core concerns about lawfulness, equal treatment, fair process, openness, accountability and adherence to the rule of law. The shift away from purely human judgment toward machine-assisted decision-making is thus more than a technical upgrade it signals a deeper change in how public authority is exercised and distributed. This paper undertakes a comparative study of AI's place in criminal justice by analyzing the legal and regulatory frameworks of India, the European Union, the United States and the United Kingdom. It asks whether algorithm-based decisions can meet established rule-of-law standards, given that such systems are often opaque, probabilistic in nature, shielded by commercial confidentiality, or built on datasets carrying historical bias. The analysis focuses closely on predictive policing, algorithmic risk assessment tools, facial recognition technology, AI-supported evidence analysis and judicial decision-support systems. The central argument advanced here is that the deeper legal risk is not simply machines displacing human decision-makers, but rather that human discretion may become obscured inside technical systems that resist scrutiny, challenge or clear lines of responsibility. Building on this analysis, the paper proposes a "Human-Centred Algorithmic Criminal Justice Model" grounded in principles of legality, proportionality, non-discrimination, transparency, explainability, human oversight, independent auditing and access to effective remedies. It contends that AI ought to function as a supporting, accountable tool in the decision-making process not as an independent source of coercive authority in matters touching on life, liberty and the right to a fair trial. The comparative review shows that each jurisdiction's approach contributes something valuable yet remains partial: the EU's rights-centred framework, the US's risk-management orientation, the UK's emphasis on transparency-based governance, and India's still-developing digital criminal justice regime. The paper closes by outlining a rights-based regulatory framework intended for India and other jurisdictions seeking to incorporate AI into criminal justice while safeguarding rule-of-law principles.