Marc Canellas
Nationally Recognized Expert in AI and Criminal Law
I am a public defender in the Forensics Division of the Maryland Office of the Public Defender, where I litigate some of the most complex criminal cases involving forensic science and emerging technologies. With a Ph.D. in aerospace engineering from Georgia Tech, a J.D. from NYU Law, and past service as a staffer in the U.S. House of Representatives, I bring a rare interdisciplinary perspective to the law—integrating rigorous technical understanding, trial and appellate litigation, and federal policy expertise.
My scholarly and professional work focuses on how software and AI systems are transforming the criminal legal system and our society, and how the law must respond to protect constitutional guarantees, civil rights, and democratic accountability.
Legal Scholarship on AI and Criminal Law
Challenging Carceral Software (job talk paper)
Applies software engineering principles to challenge software-generated suspicion and evidence in criminal cases via qualifications, reliability, search and seizure, confrontation, and due process.
Constitutional Resistance: Self-Defense Against Unlawful Federal Force under 18 U.S.C. § 111 (73 UCLA L. Rev. Discourse, 2025 forthcoming)
Arguing that 18 U.S.C. § 111, the federal statute criminalizing impeding or assaulting federal officials, has transformed into a mechanism for prosecuting civilians who resist unclear or unlawful exercises of state power, and outlines how the courts and Congress can safeguard justification defenses to prevent the erosion of constitutional limits on state violence.
Mo AI, Skidmore Problems: Governing in Our Loper Bright Era (41 J.L. & Pol., 2025, forthcoming)
Explores how the Supreme Court overturning Chevron deference undermines our ability to govern AI, and outlines recommendations for Congress and agencies to reestablish governing control.
Abolish and Reimagine: The Pseudoscience of Substance Use in the Family Regulation System (30 Geo. J. Poverty L. & Pol’y 169, 2022)
Analyzes decades of empirical literature finding that substance use is often weaponized without evidence to justify child removal, especially against marginalized parents and pregnant people—reimagining a new way to meaningfully improve child welfare.
Peer-Reviewed AI Governance Scholarship
Decoding Human Behavior with Big Data? Critical, Constructive Input from the Decision Sciences (43 AI Mag. 126, 2022) (with Konstantinos Katsikopoulos)
Challenges the premise that big data systems can reliably predict or model human decision-making. Instead, simple, transparent behavioral models can often outperform opaque AI models. (Awarded Top 10% of AI Magazine articles downloaded within 12 months of publication)
Defending IEEE Software Standards in Federal Criminal Court (54 IEEE Comput. 14, 2021)
Showing that DNA software—used in thousands of criminal cases—has evaded legitimate court and engineering scrutiny by improperly invoking scientific complexity and trade secrecy, and attacking the value of IEEE 1012 Standard for independent verification and validation. (Cited in Modern Scientific Evidence: The Law and Science of Expert Testimony, Special topics—Probabilistic software, § 30:34 (2024-2025))
Unsafe at Any Level: The U.S. NHTSA’s Levels of Automation Are a Liability for Automated Vehicles (63 Commun. ACM 31, 2020) (with Rachel Haga)
Critiquing the U.S. National Highway Traffic Safety Administration’s levels of automation framework for autonomous vehicles, arguing that it obfuscates liability and misrepresents system safety.
Lost in Translation: Building a Common Language for Regulating Autonomous Weapons (35 IEEE Technol. Soc. Mag. 50, 2016) (with Rachel Haga)
Revealing that the failure to govern autonomous weapons results not from intractable disagreement but from a lack of shared language, showing that legal norms, like accountability, autonomy, and meaningful human control must be operationalized and governed as engineering requirements.
Litigation and Expert Consulting
As a public defender I’ve managed caseloads of 180+ cases including violent felonies and been a lead juvenile defense attorney with 40+ clients. Today, I specialize in forensics litigation in complex cases, routinely challenging software and AI with respect to demanding proper discovery, challenging reliability under Daubert, and litigating violations of the Fourth Amendment, Confrontation Clause and Due Process involving forensic software, including:
- DNA software (e.g., STRmix and TrueAllele)
- AI-generated audio and images
- Geofence
- ShotSpotter
- AI ballistics matching software (e.g., Cadre)
I’ve also argued at the appellate level including an unsuccessful Daubert challenge to TrueAllele DNA software (Harvin v. State, 263 Md.App. 326 (2024), cert. denied, 489 Md. 338 (2025)) [video recording]
Training Attorneys for the Future of Law and AI
I’ve delivered invited lectures and CLEs nationwide on AI, forensic evidence, and expert testimony, including for the:
- National Institute of Standards and Technology (forthcoming)
- Santa Fe Institute (forthcoming)
- Texas Criminal Defense Lawyers Association (forthcoming)
- National Association of Criminal Defense Lawyers
- American Academy of Forensic Sciences
- Federal Communications Bar Association
- Maryland Criminal Defense Attorneys Association
- New York State Defenders Association
- Pennsylvania Association of Criminal Defense Lawyers
- The Athens Roundtable on Artificial Intelligence and the Rule of Law
- Questioning Forensics at the Legal Aid Society of New York
I also served on the American Bar Association’s AI Legal Ethics Working Group as Chair of the Oversight Committee, tasked with interpreting ABA House of Delegates Resolution 112 on the use of AI in the practice of law.
AI Policy Advocacy
I previously served as a Congressional science and technology fellow in the U.S. House of Representatives managing appropriations, policy, and media for portfolios including justice, AI, defense, surveillance, and homeland security.
Currently, I serve as an Advisory Board Member and past-Chair of the IEEE-USA AI Policy Committee, where I advise the Executive Branch and Congress on legislation and policy related to AI. Our recommendations addressing the governance of DNA software, digital forensics techniques, and forensic and surveillance tools have been cited in successful amicus briefs.
For example, in Commonwealth v. Arrington, 493 Mass. 478 (2024) the Massachusetts Supreme Judicial Court affirmed the trial court’s exclusion of defendant’s cell phone’s frequent location history as unreliable—agreeing with the Brief of Amici Curiae ACLU, ACLU of Massachusetts, EFF, NACDL, & MACDL in Support of Appellee & Affirmance which cited multiple recommendations from IEEE-USA, including my report Trustworthy Evidence for Trustworthy Technology: An Overview of Evidence for Assessing the Trustworthiness of Autonomous and Intelligent Systems, Law Committee of the IEEE Global Initiative & IEEE-USA AI Policy Committee (Sept. 29, 2022) (with Jeanna Matthews and Bruce Hedin).
Let’s chat
For collaborations, speaking engagements, or further information, please reach out or connect with me on LinkedIn.