What Factors Influence an Artificial Intelligence Case in Litigation?

The governing theory depends on the facts and relationship between parties. Contractual breach claims require proof that the AI system failed to meet performance standards, accuracy thresholds, or functional specifications outlined in the licensing or services agreement. Negligence-based theories turn on whether the AI developer or deployer failed to exercise reasonable care in system design, testing, data validation, or operational monitoring. Intellectual property claims allege unauthorized use of training data, model architecture, or algorithmic processes. Regulatory violation theories assert that the AI system's operation breached data privacy laws, algorithmic transparency rules, or sector-specific compliance frameworks. The theory selected shapes which party bears the burden of proof, what evidence is discoverable, and which affirmative defenses remain viable.

Proper service of process is a jurisdictional prerequisite; defective service can delay proceedings or provide grounds for dismissal. In New York, service must comply with CPLR Article 3 requirements, which permit service by personal delivery, mail with acknowledgment, or, in limited circumstances, electronic means if authorized by court order or statute. For corporate defendants, service often targets the registered agent or officer. Timing matters critically because the defendant's time to respond begins upon service, and any delay in establishing proper service can extend the pre-answer period, compress discovery deadlines, and create record gaps if the defendant raises service defects in a motion to dismiss.

Preservation obligations include the AI system's source code, model architecture specifications, training datasets and their metadata, version control logs, performance benchmarks, testing protocols, user documentation, internal communications discussing system design choices or known limitations, customer complaints or error reports, regulatory compliance records, and any communications with third parties regarding data licensing or model training. A party that fails to preserve material evidence faces sanctions ranging from adverse inference instructions to case dismissal. Courts recognize that AI systems generate voluminous data; however, that volume does not excuse failure to preserve documents that are relevant to the dispute. Counsel must work with technical teams to identify data custodians, establish preservation protocols, and document the chain of custody for digital evidence.

Expert testimony is nearly always necessary to establish causation, technical feasibility, industry standards, and the AI system's design or operational defects. Under Federal Rule of Civil Procedure 26(a)(2), each party must disclose expert reports containing the expert's opinions, the basis for those opinions, the facts and data considered, and the expert's qualifications. In New York state court, similar disclosure requirements apply through CPLR Article 31. Deficient expert reports may be excluded, forcing a party to proceed without technical support for its claims. Experts in artificial intelligence cases typically address machine learning methodology, data quality issues, model validation practices, algorithmic bias, system performance metrics, and industry standards for transparency and testing.

Contractual limitation clauses often shield AI vendors from liability for indirect, consequential, or punitive damages, or cap recovery to fees paid. A defendant may assert that the plaintiff failed to mitigate harm or failed to follow system documentation. Regulatory safe harbors, such as those in the Fair Housing Act or Equal Employment Opportunity laws, may protect an AI system if the defendant can show the system was designed and tested to avoid discriminatory outcomes. Lack of causation is a powerful defense; if the AI system's output did not materially contribute to the alleged harm, liability cannot attach. Some defenses require factual development during discovery; others may be resolved on a motion to dismiss if the complaint itself reveals that no viable claim exists.

A motion to dismiss under CPLR 3211 or Federal Rule 12(b)(6) tests whether the complaint states a claim upon which relief can be granted. If the plaintiff fails to allege sufficient facts about the AI system's defect, the defendant's conduct, or causation, the court may dismiss the case before discovery begins. Summary judgment under CPLR 3212 or Federal Rule 56 allows either party to eliminate disputed issues of fact if the record evidence establishes that no genuine issue remains for trial. In artificial intelligence cases, summary judgment often turns on expert testimony about causation and system performance. A defendant that can present credible expert evidence that the AI system performed as designed may prevail on summary judgment.

Upon receipt of a complaint or demand letter, a corporation should immediately issue a preservation notice to all relevant employees and departments, instruct IT to secure backup copies of all AI system code, training data, performance logs, and related communications, and consult with counsel about jurisdiction, venue, and applicable law. The corporation should gather all contracts, licensing agreements, and regulatory compliance records related to the AI system. Delaying preservation or failing to issue a timely hold notice can result in adverse inferences if evidence is later lost or destroyed. A corporation should also assess whether insurance coverage applies and notify its insurers within the timeframe specified in the policy.

New York state courts apply CPLR procedural rules and New York substantive law to artificial intelligence cases unless a federal question or diversity jurisdiction applies. In New York state court, a corporation should verify that service of process complies with CPLR Article 3 requirements and that the court has personal jurisdiction over the defendant. Discovery in New York state court proceeds under CPLR Article 31, which permits broad discovery of materials relevant to any claim or defense; however, parties may seek protective orders to limit disclosure of trade secrets or proprietary information. A corporation defending in New York state court should work with counsel to craft discovery responses that comply with CPLR timing requirements, typically 20 days from service of the demand, and should preserve all evidence that may be requested during discovery to avoid sanctions for late or incomplete production.