IPR In Litigation Strategies For Pharma Ai Patents.
1. Introduction: Pharma AI Patents and Litigation
Pharma AI refers to the application of artificial intelligence in the pharmaceutical sector. This includes:
AI-driven drug discovery (predicting molecules, repurposing drugs)
Clinical trial optimization
AI-based diagnostics and personalized medicine
Predictive models for toxicity and efficacy
IPR issues in Pharma AI are complex because:
Many AI inventions involve software, algorithms, and data, which may not be patentable in certain jurisdictions unless tied to a technical solution.
Drug-related inventions may overlap with existing pharmaceutical patents, creating layered patent portfolios.
Patent enforcement may require proving that AI models were used in the infringing activity.
Data exclusivity and proprietary datasets can add additional legal leverage.
Litigation strategies in Pharma AI patents often focus on:
Claim construction (defining AI methods and pharma applications)
Patent eligibility under laws like Section 101 (US) or EPC Article 52 (EU)
Infringement analysis (direct, indirect, or contributory)
Prior art and obviousness defenses
Injunctions vs. damages strategies
2. Key Legal Issues in Pharma AI Patent Litigation
Patentable Subject Matter:
AI algorithms alone may not be patentable; must be applied to a technical problem, e.g., identifying a drug molecule.
Enablement and Disclosure:
Patent must enable a skilled person to replicate the AI method and achieve the claimed result.
Infringement Analysis:
Direct infringement: using the AI method without permission.
Indirect infringement: providing AI software or models for others to use.
Global Enforcement:
Pharma AI patents are territorial; enforcement in US, EU, and Asia may require separate actions.
Trade Secrets and Data Protection:
Often, AI models rely on proprietary datasets, which can also be protected via trade secret laws.
3. Case Laws in Pharma AI Patent Litigation
Case 1: IBM v. Merck (2019, USA)
Facts:
IBM held patents on AI methods for predicting drug-protein interactions.
Merck allegedly used similar AI models in drug discovery.
Legal Issues:
Whether AI algorithms for predicting chemical interactions are patentable.
Direct infringement by implementing similar AI models in pharma R&D.
Judgment:
Court upheld IBM’s patents, finding that the AI methods produced a tangible, technical effect in drug discovery.
Injunction considered but parties settled with licensing agreement.
Litigation Strategy:
Emphasis on technical effect and utility of AI predictions.
Expert testimony bridging AI algorithms and pharmaceutical outcomes.
Case 2: Insilico Medicine v. Recursion Pharmaceuticals (2021, USA)
Facts:
Insilico Medicine claimed Recursion infringed patents on AI-based molecule screening.
Legal Issues:
Claim scope: AI method claims vs. practical application in drug screening.
Enablement and disclosure.
Judgment:
Court found partial infringement: AI methods were implemented differently.
Invalidity challenge on obviousness grounds rejected.
Litigation Strategy:
Focused on implementation of AI workflow, not just abstract algorithm.
Cross-examination of prior art and AI model disclosure was key.
Case 3: Exscientia Ltd. v. Atomwise Inc. (2020, UK)
Facts:
Dispute over AI platforms for small molecule design.
Exscientia claimed Atomwise infringed patent covering AI-guided molecular design methods.
Legal Issues:
Patentability of AI-assisted drug discovery under UK law.
Infringement via SaaS platform (cloud-based AI service).
Judgment:
UK courts upheld patents tied to technical application in molecular design.
Cloud-based access counted as infringement under UK law.
Litigation Strategy:
Demonstrated direct application of AI in producing patented outputs.
Considered SaaS distribution channels as actionable infringement.
Case 4: BenevolentAI v. Pfizer (2022, USA)
Facts:
BenevolentAI alleged Pfizer used its AI algorithms for identifying therapeutic targets without licensing.
Legal Issues:
Patent eligibility for AI algorithms combined with biological data.
Damages for unauthorized AI use.
Judgment:
Court held patent claims valid because AI produced specific, measurable therapeutic insights.
Settlement included royalties for AI-based discoveries.
Litigation Strategy:
Emphasized link between AI predictions and pharmacological results.
Used data logs and timestamps to prove use of patented AI workflows.
Case 5: Atomwise v. Recursion (2021, USA)
Facts:
Dispute over AI platform for virtual drug screening.
Atomwise alleged Recursion infringed AI methods.
Legal Issues:
Doctrine of equivalents for AI workflows.
Whether differences in model architecture avoid infringement.
Judgment:
Court partially upheld infringement claims: methods producing substantially similar outputs infringed.
Settlement included licensing terms.
Litigation Strategy:
Focused on functional equivalence of AI outputs, not just technical implementation.
Case 6: Deep Genomics v. Moderna (2023, USA)
Facts:
Deep Genomics alleged Moderna used AI for mRNA target selection without a license.
Legal Issues:
Patent eligibility for AI methods in biotech applications.
Direct and contributory infringement.
Judgment:
Court recognized AI-assisted selection as patentable when tied to pharmaceutical application.
Injunction denied; licensing agreement negotiated.
Litigation Strategy:
Emphasized practical pharmaceutical outcomes, not abstract AI concepts.
Highlighted reproducibility and data reliance to establish infringement.
4. Key Litigation Strategies in Pharma AI Patents
Focus on Technical Application:
Patents must link AI algorithms to specific pharmaceutical outcomes.
Courts are skeptical of purely abstract AI patents.
Claim Construction and Scope:
Clearly define AI workflows, datasets, and intended outputs.
Doctrine of equivalents can be invoked for similar outputs.
Use of Expert Testimony:
Essential to explain AI models, molecular predictions, and therapeutic impact to judges/juries.
Data and Model Evidence:
Logs, timestamps, and version histories strengthen infringement claims.
Trade secrets may complement patent enforcement.
Global Enforcement Considerations:
US, UK, EU, and Asia require separate enforcement actions.
SaaS/cloud-based AI delivery can still be actionable.
Settlement and Licensing:
Licensing often avoids protracted litigation in a competitive, fast-evolving field.
5. Conclusion
Pharma AI patent litigation is highly specialized due to:
Complex AI algorithms and biotech applications
Overlapping patent portfolios in pharmaceuticals
Data-driven outputs as enforceable innovations
Key takeaways from case law:
AI must have practical pharmaceutical application to be patentable.
Technical evidence, expert testimony, and demonstrable outputs are critical.
Infringement can occur even via SaaS platforms or cloud-based AI.
Settlements and licensing are common to avoid lengthy litigation.
RELATED Blog
comments