How AI Is Changing Pharma and Biotech Contracts in 2026

The shift is already evident in large industry deals. Two large pharmaceutical companies said in March 2026 they were extending a partnership on the discovery of new drugs in a collaboration that could be worth up to $2.75 billion, using artificial intelligence (AI) technology. The deal shows how AI capabilities are moving from experimental technology to big commercial arrangements for drug development.

This poses a new challenge for legal and commercial teams traditional pharmaceutical contracts were not written for technology that can learn from data, generate new outputs, and evolve throughout the relationship.

 

AI Is Turning Data into a Strategic Contract Asset

Data has always been a precious commodity in pharmaceutical research. Results from clinical trials, genomic data, molecular libraries, safety data, and past research may represent years of investment. That value is amplified by AI, as large datasets can be used to train models and generate new insights.

This changes the fundamental contractual question from Who owns the data? to a far larger set of questions:

• Who has access to the data?
• Can the data be used to train an AI model?
• Can the data be used to improve the model?
• Who owns the insights that are derived from the data?
• Can third parties access derived data?
• What happens to trained models after termination?
• Can the resulting model help another pharmaceutical company?
  
  

These questions are becoming more and more common in life sciences transactions. Current industry legal analysis highlights data scope, ownership, permitted use, provenance, audit rights, and restrictions on AI use as important areas for negotiation.

For example, a pharmaceutical company might hire a CRO to study its historic clinical data. If the CRO is using an external AI platform, the contract should cover whether that information can be retained, used to train the model, or integrated into a wider commercial AI system.

This distinction is important because the deletion of the original dataset does not mean that the influence that data has already had on a machine-learning model is deleted.

AI Model Licensing Is Creating New Contract Structures

Not all AI partnerships are created equal in life sciences. One model is that the AI company itself does the analysis and delivers the research results to the pharma company. Another is to license the AI model directly to the pharma company. A third involves several organizations contributing data to a shared AI system. Legal risks are based on the structure.

For example, in a straight licensing arrangement between a model and the pharmaceutical company, the company may want expansive rights to use the AI-generated results in drug development, regulatory filings, and commercialization. Meanwhile, the AI company might want to protect its underlying model, training data, and proprietary technology. This creates a critical contractual distinction between the AI model and the outputs the model generates.

Ownership and permitted use may need to be defined separately in a contract:

1. Background intellectual property
2. Training Data
3. AI models
4. Upgrades to the model
5. Outputs produced by AI
6. Drug Targets & Candidates
7. Patents & Inventions
8. Clinical & Regulatory Data
   
   

These distinctions are increasingly being factored into todays life sciences deals, as AI-native biotech companies license specialist models and platforms to pharma.

 

Who Owns an AI-Generated Drug Discovery?

One of the most complex questions that arise from AI-enabled research is ownership of the resulting discovery. Imagine a biotech company teaming up with a pharma manufacturer.

The biotech company is providing an AI platform. The company has decades of proprietary research data in pharmaceuticals. The AI system identifies a new target and suggests several molecules.

Who owns the resulting invention?

The answer cannot safely be left to a generic intellectual property clause. The agreement might need to delineate between the background IP of the parties, the AI platform, the data that is input into the platform, the outputs of the model, and the therapeutic candidate that is ultimately developed from those outputs.

The parties may also have to think about whether the AI-generated discoveries are patentable, who controls patent applications, and how rights will be split if both parties contributed to the resulting invention. These questions are especially important when the AI platform is used across several pharmaceutical collaborations.

 

 

Training Rights Are Becoming a Major Negotiation Point

A company may agree to share proprietary data for a specific research project without realizing that the data can be used to improve the suppliers AI model. That difference can have important commercial implications.

Say a biotech company is giving an AI provider access to thousands of its proprietary molecular frameworks. The AI provider then uses those structures to improve its model. Next, the improved model is used by another pharmaceutical company. Even if the original molecular files are not shared, the first biotech company may wonder if its contribution has created commercial value for a competitor.

Therefore, contracts increasingly need to include explicit provisions for AI training and model improvement. Parties may agree to limit:

• Training models with confidential data
• Data retention after the project
• Use of data for other research purposes
• Sharing insights gathered
• Using the data of competing products
• Custom training of models with proprietary data
• Joining the data with third-party data sets
  
  

In some cases, companies may block public AI tools from being used on sensitive information but may allow the use of enterprise AI tools provided that they are safeguarded with cybersecurity, confidentiality, and intellectual property protections.

 

AI Is Bringing Regulatory Obligations into Commercial Contracts

AI regulation is also changing contracts in pharma. In January 2026, the U.S. Food and Drug Administration and European Medicines Agency jointly identified ten guiding principles for good AI practice in drug development. The principles include human-centric design, risk-based approaches, data governance, documentation, model performance assessment, and lifecycle management.

The importance of contracting is great. A pharmaceutical company that outsources AI-enabled research may need contractual assurances that the technology is validated, documented, and appropriately monitored.

Contracts may thus include obligations in relation to:

• Governance of Data
• Verification of Model
• Human supervision
• Performance monitoring
• Documentation
• Risk Management
• Cooperation on regulation
• Management of change
• Right to audit
  
  

This is especially true when AI is used in clinical trials, safety monitoring, or other activities that may ultimately be used to support regulatory submissions.

 

Confidentiality Clauses Need an AI Upgrade

Old confidentiality clauses may not apply to the way new artificial intelligence systems process information.