Marc Canellas (Maryland Office of The Public Defender) has posted Machine-Generated Evidence on SSRN. Here is the abstract:
Our legal system is being rewritten in code. In the criminal legal system, from AI-authored police reports to face recognition and DNA software, carceral machines–systems built from hardware, software and AI–now generate the suspicion, evidence, and convictions. This transformation has raised one of the most urgent legal questions of our time: how should constitutional rights, evidentiary and discovery rules apply to machine-generated evidence?
Courts and lawmakers are beginning to take notice. In 2023, Supreme Court Chief Justice John Roberts called for “caution and humility” as AI reshapes the law. In 2025, the Federal Judicial Conference proposed a new Rule of Evidence 707 affirming that “machine-generated evidence” must meet existing reliability standards. But courts are still left to wonder what caution, humility, rights or rules mean for these machines that even most purported “experts” barely understand.
This Article offers the first comprehensive framework for that task. Drawing on expertise in AI, software engineering, and human-machine interaction, it first identifies the machine-law mismatch: where courts assign legal responsibility to human actors like police, affiants, and forensic analysts, despite machines increasingly controlling the decisions. Too often human actors lack the access or expertise to evaluate or control the machines they rely on, leading to violations of law and wrongful convictions.
The solution is to rigorously apply our rights and rules to machines as they are. This Article walks through the doctrines of the criminal adjudication process to show how search and seizure, discovery, confrontation, and expert qualifications and reliability should be applied to machines. Courts must treat machine-generated evidence the way engineers do: through independent verification and validation (IV&V) conducted and evaluated by software experts. Only then can courts fulfill their constitutional and statutory gatekeeping role to ensure that all evidence, whether generated by humans or machines, remain subject to the rule of law.
Recommended.