Regulon — MD-HB-712 · MD HB 712 (AI Product Liability)

WHAT THIS BILL REGULATES · 3 REQUIREMENT TYPES

How Is This Bill Enforced

Enforcement Authority

Private right of action for any person harmed by a dangerous or defective AI product. The Attorney General may also bring an action against a developer or deployer for harm caused by a dangerous or defective product; before seeking injunctive relief, the AG must provide 30 days' written notice and, when possible, an opportunity to cure.

Private Right of Action

Private right of action for any person harmed by a dangerous or defective AI product.

Penalties

Private plaintiffs may recover compensatory damages for harm (property damage, personal injury, financial or reputational injury, emotional harm, loss of consortium, or death). Recovery is reduced by the percentage of fault attributable to the plaintiff. The Attorney General may obtain injunctive relief, civil penalties (capped at the applicable noneconomic damages limitation under § 3–2A–09 in cases involving personal injury or death), damages, restitution, other compensation on behalf of harmed individuals, and costs of the action. Joint and several liability applies to the portion of harm attributable to the developer and deployer.

Verbatim statutory text on the left; plain-language analysis and a per-section checklist on the right. Numbered markers cross-link to the matching checklist row.

Statutory Text

Analysis & Obligations

Md. Code, Cts. & Jud. Proc. § 3–2701

Definitions

(A)–(O) IN THIS SUBTITLE THE FOLLOWING WORDS HAVE THE MEANINGS INDICATED. (B) "ARTIFICIAL INTELLIGENCE SYSTEM" MEANS AN ENGINEERED OR MACHINE–BASED SYSTEM WITH VARYING LEVELS OF AUTONOMY THAT CAN, FOR EXPLICIT OR IMPLICIT OBJECTIVES, INFER FROM THE INPUTS IT RECEIVES HOW TO GENERATE OUTPUTS THAT CAN INFLUENCE PHYSICAL OR VIRTUAL ENVIRONMENTS. (C) "CONSEQUENTIAL DECISION" MEANS A DECISION THAT HAS A LEGAL OR SIMILARLY SIGNIFICANT EFFECT ON AN INDIVIDUAL'S ACCESS TO THE CRIMINAL JUSTICE SYSTEM, HOUSING, EMPLOYMENT, CREDIT, EDUCATION, HEALTH CARE, OR INSURANCE. (D) (1) "DEPLOYER" MEANS A PERSON, INCLUDING A DEVELOPER, THAT USES OR OPERATES AN ARTIFICIAL INTELLIGENCE SYSTEM FOR USE BY THE DEPLOYER OR THIRD PARTIES. (2) "DEPLOYER" DOES NOT INCLUDE AN INDIVIDUAL OR SMALL BUSINESS THAT HAS FEWER THAN: (I) 20 EMPLOYEES; OR (II) 10,000 USERS OF ITS PRODUCT. (E) (1) "DEVELOPER" MEANS A PERSON THAT DESIGNS, CODES, PRODUCES, OWNS, OR SUBSTANTIALLY MODIFIES AN ARTIFICIAL INTELLIGENCE SYSTEM FOR USE BY A DEVELOPER OR FOR USE BY THIRD PARTIES. (2) "DEVELOPER" DOES NOT INCLUDE A PERSON THAT USES AN OPEN SOURCE ARTIFICIAL INTELLIGENCE SYSTEM AND DOES NOT SUBSTANTIALLY MODIFY THE ARTIFICIAL INTELLIGENCE SYSTEM. (F) "DESIGN" MEANS THE INTENDED OR KNOWN PHYSICAL AND MATERIAL CHARACTERISTICS OF A PRODUCT, INCLUDING: (1) ANY INTENDED OR KNOWN FORMULATION OR CONTENT OF THE PRODUCT; AND (2) THE USUAL RESULT OF THE INTENDED METHOD OF DEVELOPMENT OR OTHER PROCESS USED TO PRODUCE THE PRODUCT, INCLUDING UNEXPECTED SKILLS OR BEHAVIORS THAT APPEAR IN A PRODUCT. (G) "EXPRESS WARRANTY" MEANS ANY MATERIAL, POSITIVE STATEMENT, AFFIRMATION OF FACT, PROMISE, OR DESCRIPTION RELATING TO A PRODUCT, INCLUDING A SAMPLE OR MODEL OF A PRODUCT. (H) "GENERATIVE ARTIFICIAL INTELLIGENCE SYSTEM" MEANS AN ARTIFICIAL INTELLIGENCE SYSTEM THAT CAN GENERATE DERIVED SYNTHETIC CONTENT, SUCH AS TEXT, IMAGES, VIDEO, AND AUDIO, THAT EMULATES THE STRUCTURE AND CHARACTERISTICS OF THE ARTIFICIAL INTELLIGENCE SYSTEM'S TRAINING DATA. (I) "HARM" MEANS: (1) DAMAGE TO PROPERTY OTHER THAN THE ARTIFICIAL INTELLIGENCE SYSTEM ITSELF; (2) PERSONAL PHYSICAL, FINANCIAL, OR REPUTATIONAL INJURY, ILLNESS, OR DEATH; (3) MENTAL OR PHYSIOLOGICAL ANGUISH, EMOTIONAL HARM, OR DISTORTION OF A PERSON'S BEHAVIOR THAT WOULD BE HIGHLY OFFENSIVE TO A REASONABLE PERSON; OR (4) ANY LOSS OF CONSORTIUM OR SERVICES, OR OTHER LOSS DERIVING FROM ANY TYPE OF HARM DESCRIBED UNDER THIS SUBSECTION. (J) "HIGH–IMPACT ARTIFICIAL INTELLIGENCE SYSTEM" MEANS ANY ARTIFICIAL INTELLIGENCE SYSTEM, REGARDLESS OF THE NUMBER OF PARAMETERS AND SUPERVISION STRUCTURE, THAT: (1) IS USED, OR REASONABLY FORESEEABLY MAY BE USED, AS A CONTROLLING FACTOR IN MAKING CONSEQUENTIAL DECISIONS; (2) IS USED, OR REASONABLY FORESEEABLY MAY BE USED, TO CATEGORIZE GROUPS OF PERSONS BY PROTECTED CHARACTERISTICS, SUCH AS RACE, ETHNIC ORIGIN, OR RELIGIOUS BELIEF; (3) IS USED, OR REASONABLY FORESEEABLY MAY BE USED, IN THE DIRECT MANAGEMENT OR OPERATION OF CRITICAL INFRASTRUCTURE; (4) IS USED, OR REASONABLY FORESEEABLY MAY BE USED, IN A VEHICLE, IN A MEDICAL DEVICE, OR IN THE SAFETY SYSTEM OF A VEHICLE OR MEDICAL DEVICE; (5) IS USED, OR REASONABLY FORESEEABLY MAY BE USED, TO ENGAGE IN A SYNTHETIC RELATIONSHIP; OR (6) EXHIBITS, OR COULD BE EASILY MODIFIED TO EXHIBIT, HIGH LEVELS OF PERFORMANCE AT TASKS THAT POSE A SERIOUS RISK TO ECONOMIC SECURITY OR PUBLIC HEALTH OR SAFETY. (K) (1) "MATERIAL FACT" MEANS ANY SPECIFIC CHARACTERISTIC OR QUALITY OF THE PRODUCT. (2) "MATERIAL FACT" DOES NOT INCLUDE A GENERAL OPINION ABOUT THE PRODUCT OR ITS QUALITY. (L) "METHOD OF DEVELOPMENT" MEANS THE SELECTION OF TRAINING DATA FOR THE PRODUCT, INCLUDING THE TRAINING, TESTING, AUDITING, AND FINE–TUNING OF THE PRODUCT. (M) "PERSON" MEANS ANY INDIVIDUAL, CORPORATION, COMPANY, ASSOCIATION, FIRM, PARTNERSHIP, SOCIETY, JOINT STOCK COMPANY, OR ANY OTHER ENTITY, INCLUDING ANY GOVERNMENTAL ENTITY OR UNINCORPORATED ASSOCIATION OF PERSONS. (N) "PRODUCT" MEANS A HIGH–IMPACT ARTIFICIAL INTELLIGENCE SYSTEM OR A GENERATIVE ARTIFICIAL INTELLIGENCE SYSTEM. (O) "SYNTHETIC RELATIONSHIP" MEANS A SERIES OF INTERACTIONS BETWEEN AN INDIVIDUAL AND AN ARTIFICIAL INTELLIGENCE SYSTEM THAT MIMICS HUMAN INTERACTION AND EMOTIONAL RESPONSES.

Section 3–2701 establishes the defined terms used throughout the new AI product liability subtitle. Key features include a broad definition of artificial intelligence system tracking the NIST framework language, a six-prong high-impact artificial intelligence system definition that sweeps in consequential decision-making, biometric classification, critical infrastructure, vehicles and medical devices, synthetic relationships, and frontier-capable systems. The bill defines product to mean either a high-impact AI system or a generative AI system, setting the scope of all causes of action. Notably, design is defined to include unexpected skills or behaviors that appear in a product, explicitly bringing emergent AI capabilities within the design-defect framework. The deployer definition carves out small businesses with fewer than 20 employees or 10,000 users, and developer excludes unmodified open-source users.

Md. Code, Cts. & Jud. Proc. § 3–2702

Exclusion for peer-reviewed scientific research

THIS SUBTITLE DOES NOT APPLY TO ARTIFICIAL INTELLIGENCE SYSTEMS USED ONLY FOR PEER–REVIEWED SCIENTIFIC RESEARCH.

Section 3–2702 carves out AI systems used solely for peer-reviewed scientific research from the entire subtitle. This exemption is absolute — if the AI system is used only for that purpose, no cause of action under this subtitle applies.

Md. Code, Cts. & Jud. Proc. § 3–2703

Causes of action and rebuttable presumptions

DeployerDeveloper

(A) 1 A PERSON MAY BRING AN ACTION IN ACCORDANCE WITH THIS SECTION AGAINST A DEVELOPER OR DEPLOYER FOR HARM CAUSED BY A PRODUCT THAT IS DANGEROUS OR DEFECTIVE.

(B) 2 IN AN ACTION AGAINST A DEVELOPER ALLEGING HARM CAUSED BY A DANGEROUS OR DEFECTIVE PRODUCT DUE TO THE DEFECTIVE DESIGN OF THE PRODUCT, THE PLAINTIFF SHALL PROVE BY A PREPONDERANCE OF THE EVIDENCE THAT THE HARM WAS PROXIMATELY CAUSED BY THE FAILURE OF THE DEVELOPER TO EXERCISE REASONABLE CARE AND THAT, AT THE TIME THE PRODUCT LEFT THE DEVELOPER'S CONTROL: (1) THE DEVELOPER KNEW OR REASONABLY SHOULD HAVE KNOWN OF THE DESIGN DEFECT THAT CAUSED THE HARM; (2) THE DEVELOPER ACCOUNTED FOR INTENDED USES AND REASONABLY FORESEEABLE UNINTENDED USES OF THE PRODUCT; AND (3) A TECHNOLOGICALLY FEASIBLE AND PRACTICAL ALTERNATIVE DESIGN EXISTED THAT WOULD HAVE MITIGATED OR AVOIDED THE FORESEEABLE RISK OF HARM WITHOUT SIGNIFICANTLY IMPAIRING THE INTENDED USE OF THE PRODUCT.

(C)(1)–(3) 3 IN THIS SUBSECTION, "ADEQUATE INSTRUCTION OR WARNING" MEANS AN INSTRUCTION OR A WARNING REGARDING A DANGER THAT WOULD SUFFICIENTLY INFORM A REASONABLY PRUDENT PERSON ON THE SAFE USE AND DANGERS OF A PRODUCT. (2) IN AN ACTION AGAINST A DEVELOPER ALLEGING HARM CAUSED BY A DANGEROUS OR DEFECTIVE PRODUCT DUE TO A FAILURE TO PROVIDE ADEQUATE INSTRUCTION OR WARNING, A PLAINTIFF SHALL PROVE BY A PREPONDERANCE OF THE EVIDENCE THAT THE HARM WAS PROXIMATELY CAUSED BY THE FAILURE TO PROVIDE ADEQUATE INSTRUCTION OR WARNING AND THAT, AT THE TIME THE PRODUCT LEFT THE DEVELOPER'S CONTROL, THE DEVELOPER KNEW OR REASONABLY SHOULD HAVE KNOWN OF THE DANGER POSED BY THE PRODUCT. (3) IT IS A DEFENSE TO AN ACTION AGAINST A DEVELOPER FOR HARM CAUSED BY A FAILURE TO INSTRUCT OR WARN THAT: (I) THE USER OR CONSUMER OF THE PRODUCT WAS AT LEAST 17 YEARS OLD; AND (II) THE PRODUCT DANGER WAS KNOWN OR OPEN AND OBVIOUS TO THE USER OR CONSUMER OF THE PRODUCT, OR SHOULD HAVE BEEN KNOWN OR OPEN AND OBVIOUS TO THE USER OR CONSUMER OF THE PRODUCT.

(D) 4 IN AN ACTION AGAINST A DEVELOPER ALLEGING HARM CAUSED BY A DANGEROUS OR DEFECTIVE PRODUCT DUE TO THE FAILURE OF THE PRODUCT TO CONFORM TO AN EXPRESS WARRANTY, THE PLAINTIFF SHALL PROVE BY A PREPONDERANCE OF THE EVIDENCE THAT THE HARM WAS PROXIMATELY CAUSED BY THE FAILURE OF THE DEVELOPER TO EXERCISE REASONABLE CARE AND THAT: (1) THE PLAINTIFF REASONABLY RELIED ON AN EXPRESS WARRANTY MADE BY A DEVELOPER ABOUT A MATERIAL FACT CONCERNING THE SAFETY OF THE PRODUCT; AND (2) (I) 1. THE PRODUCT FAILED TO CONFORM TO THE EXPRESS WARRANTY; AND 2. THE FAILURE OF THE PRODUCT TO CONFORM TO THE EXPRESS WARRANTY RESULTED IN THE HARM; OR (II) 1. THE EXPRESS WARRANTY WAS UNTRUE; AND 2. HAD THE EXPRESS WARRANTY BEEN TRUE, THE HARM WOULD NOT HAVE OCCURRED.

(E)(1)–(3) 5 A DEPLOYER MAY BE HELD LIABLE IN PLACE OF A DEVELOPER IN AN ACTION UNDER THIS SECTION IF THE DEPLOYER: (I) MADE A MATERIAL AND SUBSTANTIAL CHANGE TO THE PRODUCT; OR (II) INTENTIONALLY MISUSED THE PRODUCT CONTRARY TO AN EXPRESS WARRANTY AND THAT MISUSE WAS THE PROXIMATE CAUSE OF HARM SUFFERED BY A PLAINTIFF. (2) (I) USE OF A PRODUCT THAT IS INTENDED BY THE DEVELOPER OF THE PRODUCT DOES NOT CONSTITUTE A MISUSE OR MATERIAL OR SUBSTANTIAL CHANGE OF THE PRODUCT UNDER THIS SECTION. (II) IF A DEVELOPER DOES NOT SPECIFY AN INTENDED USE FOR THE PRODUCT, INTENDED USE MAY BE INFERRED BY THE TARGETED MARKET AND MANNER OF DISTRIBUTION OF THE PRODUCT. (3) A DEPLOYER LICENSING A PRODUCT MAY NOT BE HELD LIABLE TO A PLAINTIFF FOR A VIOLATION UNDER THIS SECTION SOLELY BY REASON OF OWNERSHIP OR USE OF A PRODUCT.

(F)(1) 6 IN AN ACTION BROUGHT UNDER THIS SECTION, THERE IS A REBUTTABLE PRESUMPTION THAT A PRODUCT IS NOT DANGEROUS OR DEFECTIVE IF: (1) IN AN ACTION AGAINST A DEVELOPER, THE DEVELOPER: (I) CONDUCTED A DOCUMENTED TESTING, EVALUATION, VERIFICATION, VALIDATION, AND AUDITING OF THE PRODUCT CONSISTENT WITH INDUSTRY BEST PRACTICES; (II) MITIGATED FORESEEABLE RISKS TO THE EXTENT POSSIBLE; (III) DISCLOSED FORESEEABLE RISKS AND MITIGATION TACTICS DIRECTLY TO DEPLOYERS AND CONSUMERS; AND (IV) IF THE PRODUCT IS DESIGNED FOR OR REASONABLY LIKELY TO BE USED BY MINORS, CONSIDERED AND MITIGATED FORESEEABLE RISKS TO MINORS USING THE PRODUCT, INCLUDING: 1. THE IMPACT OF THE PRODUCT ON COGNITIVE AND EMOTIONAL DEVELOPMENT; 2. THE IMPLEMENTATION OF AGE AND CONTENT RESTRICTIONS; AND 3. PROVIDING CLEAR, ACCESSIBLE DISCLOSURES TO DEPLOYERS, CONSUMERS, AND GUARDIANS ABOUT POTENTIAL RISKS;

(F)(2) 7 IN AN ACTION AGAINST A DEPLOYER, THE DEPLOYER IMPLEMENTED AND ADHERED TO A RISK MANAGEMENT POLICY THAT: (I) SPECIFIES THE POLICY BY WHICH THE DEPLOYER WILL IDENTIFY, DOCUMENT, AND MITIGATE ANY RISK, ESPECIALLY TO MINOR USERS, THAT IS REASONABLY FORESEEABLE; (II) IS CONSISTENT WITH INDUSTRY BEST PRACTICES; (III) IS REASONABLE IN LIGHT OF: 1. THE SIZE AND COMPLEXITY OF THE DEPLOYER; 2. THE NATURE AND SCOPE OF THE PRODUCT, INCLUDING THE INTENDED USES AND UNINTENDED USES, AS WELL AS ANY MODIFICATION MADE TO THE SYSTEM BY THE DEPLOYER; AND 3. THE DATA THAT THE SYSTEM, ONCE DEPLOYED, PROCESSES AS INPUTS; AND (IV) IS ELECTRONICALLY AVAILABLE TO ITS EMPLOYEES AND TO THE ATTORNEY GENERAL ON REQUEST.

Section 3–2703 is the operative heart of the bill, establishing three distinct causes of action against developers — defective design, failure to warn, and breach of express warranty — and specifying when deployers may be sued in lieu of developers. Each cause of action requires proof by a preponderance of the evidence that harm was proximately caused by the developer's failure to exercise reasonable care.

The defective-design claim requires showing the developer knew or should have known of the defect, accounted for intended and foreseeable unintended uses, and that a feasible alternative design existed. The failure-to-warn claim requires showing the developer knew or should have known of the product's danger, with a defense for adult users (≥17) when the danger was open and obvious. The express-warranty claim requires reasonable reliance on a material safety statement by the developer.

Deployer liability is limited to cases where the deployer materially and substantially changed the product or intentionally misused it contrary to an express warranty. A deployer merely licensing and using the product cannot be held liable solely on that basis. The section creates powerful rebuttable presumptions of non-defectiveness: for developers who conducted documented testing consistent with industry best practices, mitigated foreseeable risks, disclosed risks to deployers and consumers, and (for minor-facing products) addressed risks to minors; and for deployers who implemented and adhered to a risk management policy meeting specified criteria. The deployer's risk management policy must be electronically available to employees and to the Attorney General on request.

Compliance actions 5 items

Developers must exercise reasonable care in the design of their AI product, including knowing or discovering design defects, accounting for intended and reasonably foreseeable unintended uses, and ensuring no technologically feasible and practical alternative design would have mitigated the foreseeable risk of harm without significantly impairing the product's intended use.

S-01.1

Developers must provide adequate instructions or warnings regarding foreseeable dangers of the AI product, sufficient to inform a reasonably prudent person of the safe use and dangers. A defense exists for adult users (≥17) when the danger was open and obvious.

S-02.10

Developers are liable for harm caused when a product fails to conform to an express warranty about a material safety fact on which the plaintiff reasonably relied, or when the express warranty was untrue and the harm would not have occurred had it been true.

—

Developers may establish a rebuttable presumption of non-defectiveness by (1) conducting documented testing, evaluation, verification, validation, and auditing consistent with industry best practices, (2) mitigating foreseeable risks to the extent possible, (3) disclosing foreseeable risks and mitigation tactics directly to deployers and consumers, and (4) for products designed for or reasonably likely to be used by minors, considering and mitigating risks to minors including cognitive and emotional development impacts, implementing age and content restrictions, and providing clear disclosures to deployers, consumers, and guardians.

S-01.1

Deployers may establish a rebuttable presumption of non-defectiveness by implementing and adhering to a risk management policy that (1) specifies how the deployer will identify, document, and mitigate foreseeable risks, especially to minor users, (2) is consistent with industry best practices, (3) is reasonable in light of the deployer's size, complexity, the product's nature and scope, and the data the system processes, and (4) is electronically available to employees and to the Attorney General on request.

G-01.1

Md. Code, Cts. & Jud. Proc. § 3–2704

Attorney General enforcement

(A)–(D) THE ATTORNEY GENERAL MAY BRING AN ACTION AGAINST A DEVELOPER OR A DEPLOYER FOR HARM CAUSED BY A DANGEROUS OR DEFECTIVE PRODUCT DESCRIBED IN § 3–2703 OF THIS SUBTITLE. (B) (1) IN AN ACTION BROUGHT UNDER THIS SECTION, THE ATTORNEY GENERAL MAY OBTAIN: (I) INJUNCTIVE RELIEF; (II) A CIVIL PENALTY; (III) DAMAGES, RESTITUTION, OR OTHER COMPENSATION ON BEHALF OF AN INDIVIDUAL WHO HAS BEEN HARMED BY AN ARTIFICIAL INTELLIGENCE SYSTEM; AND (IV) ANY OTHER RELIEF THAT THE ATTORNEY GENERAL CONSIDERS APPROPRIATE. (2) IF THE USE OF THE SOFTWARE CAUSES PERSONAL INJURY OR DEATH, A CIVIL PENALTY IMPOSED UNDER THIS SUBSECTION MAY NOT EXCEED THE APPLICABLE LIMITATION ON NONECONOMIC DAMAGES UNDER § 3–2A–09 OF THIS TITLE. (C) IN AN ACTION BROUGHT BY THE ATTORNEY GENERAL UNDER THIS SECTION, THE ATTORNEY GENERAL IS ENTITLED TO RECOVER THE COSTS OF THE ACTION FOR THE USE OF THE STATE. (D) BEFORE BRINGING AN ACTION UNDER THIS SECTION SEEKING INJUNCTIVE RELIEF, THE ATTORNEY GENERAL SHALL PROVIDE WRITTEN NOTICE TO A DEVELOPER OR DEPLOYER AT LEAST 30 DAYS BEFORE COMMENCEMENT OF THE ACTION THAT INCLUDES A DETAILED DESCRIPTION OF THE BASIS OF THE CLAIM AND, WHEN POSSIBLE, THE OPPORTUNITY TO CURE ANY CONDITION GIVING RISE TO THE CLAIM.

Section 3–2704 authorizes the Attorney General to bring enforcement actions against developers and deployers for harm caused by dangerous or defective AI products. The AG may obtain injunctive relief, civil penalties (capped at the applicable noneconomic damages limitation under § 3–2A–09 in cases involving personal injury or death), compensatory damages and restitution on behalf of harmed individuals, and any other appropriate relief. The AG is entitled to recover costs of the action. Before seeking injunctive relief, the AG must provide 30 days' written notice with a detailed description of the claim and, when possible, an opportunity to cure.

Md. Code, Cts. & Jud. Proc. § 3–2705

Defenses and joint liability

(A)(1)–(2) CONTRIBUTORY NEGLIGENCE IS NOT A DEFENSE TO AN ACTION BROUGHT UNDER THIS SUBTITLE AND RECOVERY BY A PLAINTIFF MAY NOT BE BARRED REGARDLESS OF THE PLAINTIFF'S DEGREE OF FAULT. (2) RECOVERY BY A PLAINTIFF UNDER THIS SUBTITLE SHALL BE REDUCED IN PROPORTION TO THE PERCENTAGE OF FAULT ATTRIBUTABLE TO THE PLAINTIFF.

(B) A DEVELOPER AND A DEPLOYER MAY BE HELD JOINTLY AND SEVERALLY LIABLE FOR THE PORTION OF HARM TO THE PLAINTIFF ATTRIBUTED TO THE DEVELOPER AND DEPLOYER.

(C) 8 IT IS NOT A DEFENSE TO AN ACTION BROUGHT UNDER THIS SUBTITLE THAT AN ARTIFICIAL INTELLIGENCE SYSTEM AUTONOMOUSLY CAUSED THE HARM TO THE PLAINTIFF.

Section 3–2705 establishes three procedural rules shaping litigation under this subtitle. Contributory negligence is abolished as a complete defense — a plaintiff's recovery cannot be barred regardless of fault, but damages are reduced proportionally by the plaintiff's percentage of fault (modified comparative fault). Developers and deployers may be held jointly and severally liable for the portion of harm attributed to them. Finally, the bill explicitly forecloses the defense that an AI system autonomously caused the harm, ensuring that developers and deployers cannot disclaim liability by pointing to the AI system's independent decision-making.

Md. Code, Cts. & Jud. Proc. § 3–2706

Relationship to existing law

(1)–(2) THIS SUBTITLE: (1) SUPPLEMENTS ANY COMMON LAW TORT LIABILITY CAUSE OF ACTION AND ANY STATE PRODUCT LIABILITY LAWS; AND (2) MAY NOT BE INTERPRETED TO PROHIBIT ANY PRODUCT LIABILITY CAUSE OF ACTION INVOLVING A PRODUCT BROUGHT UNDER A DIFFERENT CLAIM UNDER PRODUCT LIABILITY COMMON LAW OR STATUTE.

Section 3–2706 is a savings clause providing that the new subtitle supplements — rather than displaces — existing common law tort liability and state product liability laws. No existing product liability cause of action is preempted by this subtitle.

Safety Harmful content prevention · self-harm escalation

3 items

S-01.1