Intellectual Property Due Diligence For Ai.
Intellectual Property Due Diligence for AI
1. Conceptual Overview
IP Due Diligence is the systematic assessment of intellectual property assets to identify:
Ownership rights
Validity of protection
Risk of infringement or litigation
Commercial value
Freedom-to-operate (FTO) and licensing obligations
For AI systems, IP due diligence is complex because AI involves:
Algorithms and software code
Training datasets (often proprietary)
Model architectures and weights
Cloud infrastructure / deployment methods
APIs and integration tools
Generated content outputs
The goal of IP due diligence in AI is to ensure that investments, acquisitions, or partnerships do not infringe on others’ IP and that the AI assets themselves are legally protected.
2. Key IP Due Diligence Areas for AI
Patent Landscape Analysis
Identify AI-related patents: models, training methods, inference optimization.
Check for overlapping patents (risk of litigation).
Copyright Assessment
AI source code, model documentation, user manuals.
Ensure licenses for third-party libraries (Open Source or commercial).
Trade Secret Review
Datasets, model weights, fine-tuning methods.
NDAs, employee agreements, cloud storage policies.
Trademark and Branding
AI system branding, logos, and product names.
Ensure no conflicts for SaaS or AI-as-a-Service offerings.
Licensing and Third-Party Dependencies
Check open-source libraries used in model training.
API contracts and data licensing terms.
Regulatory Compliance
GDPR, HIPAA, or other privacy laws affecting datasets.
Impact of legal restrictions on commercial AI deployment.
3. Case Laws Involving AI IP Due Diligence
Here are more than 5 cases demonstrating how IP due diligence (or the lack thereof) has led to litigation or influenced transactions in AI-related technology.
Case 1: Waymo LLC v. Uber Technologies Inc. (2017)
Facts:
Waymo sued Uber alleging that Uber stole AI trade secrets for autonomous driving.
Former employee Anthony Levandowski allegedly downloaded 14,000 files, including LiDAR algorithms and model designs, before joining Uber.
IP Due Diligence Angle:
Uber claimed it conducted due diligence during acquisition but failed to detect prior misappropriation.
Highlighted importance of reviewing employee movement history, prior IP ownership, and cloud-based data security.
Outcome:
Settlement with Uber transferring equity and payment.
Emphasized AI trade secrets in IP due diligence during acquisitions or partnerships.
Case 2: Epic Systems Corp. v. Tata Consultancy Services (2016–2018)
Facts:
TCS allegedly accessed Epic’s cloud-hosted healthcare software and used it to create competing AI analytics tools.
IP Due Diligence Angle:
Demonstrated failure to audit cloud access and employee compliance.
Companies need due diligence on:
Licenses for data usage
Employee activity and prior obligations
Contractual rights over AI software
Outcome:
Jury awarded over $900 million.
Reinforced IP diligence importance for both acquisitions and outsourcing of AI development.
Case 3: Tesla, Inc. v. Zoox, Inc. (2019)
Facts:
Tesla sued Zoox for theft of AI logistics and manufacturing software.
Ex-employees allegedly transferred cloud-hosted AI models to Zoox.
IP Due Diligence Angle:
During acquisition or hiring, Zoox did not verify prior employee access rights.
IP due diligence for AI acquisitions requires:
Investigating prior employers’ trade secrets
Conducting forensic audits of code transfers
Outcome:
Settlement with injunctions.
Case shows pre-acquisition IP review is crucial for cloud AI systems.
Case 4: Microsoft v. Motorola Mobility (2012–2014)
Facts:
Dispute over patents and standards-essential IP in software for mobile devices and AI-driven services.
Motorola tried to charge royalties for patents embedded in AI features of Windows Phone.
IP Due Diligence Angle:
Microsoft had to conduct thorough patent due diligence before licensing standards-essential technologies.
AI product developers must:
Map potential patent risks
Ensure proper FTO
Avoid infringement claims in AI-enabled software
Outcome:
Court imposed reasonable and non-discriminatory (RAND) licensing.
Reinforces need for patent landscape review for AI products.
Case 5: IBM Watson Health Data Dispute (2019)
Facts:
IBM Watson Health faced allegations of using proprietary datasets without proper licensing to train AI models for healthcare insights.
IP Due Diligence Angle:
Highlighted the need for dataset licensing checks.
AI due diligence must cover:
Ownership and rights of datasets
Third-party data license compliance
Data privacy regulations impacting IP ownership
Outcome:
Legal scrutiny led to IBM tightening dataset sourcing policies.
Example of IP due diligence beyond patents—covering data ownership in AI systems.
Case 6: OpenAI Partnership Dispute (Confidential Settlements)
Facts:
Early AI partnerships resulted in disputes over code, model outputs, and licensing rights.
Conflict arose over who owned derived models and API products.
IP Due Diligence Angle:
Investors and partners conducted post-funding IP due diligence.
AI IP diligence should:
Review licensing of pre-trained models
Clarify rights over derivative works
Ensure contracts clearly define ownership of outputs
Outcome:
Settlements and clarified contracts
Sets precedent for due diligence in AI model partnerships
Case 7: DeepMind/Alphabet IP Licensing Audit (2018)
Facts:
DeepMind licensed AI algorithms to healthcare providers (NHS).
Later disputes emerged over AI IP rights in derivative software.
IP Due Diligence Angle:
Demonstrates importance of:
Clear IP ownership clauses
Audit rights for cloud-hosted AI services
Verification of third-party data and software compliance
Outcome:
Contracts clarified IP rights
Reinforces role of due diligence for AI cloud-based services
4. Key Lessons from AI IP Due Diligence Cases
Trade secrets in AI: Model architecture, training data, and code are high-risk IP assets.
Employee vetting: Checking prior employer obligations prevents misappropriation.
Cloud oversight: Audit cloud access logs and file transfers.
Dataset licensing: Verify that all training data complies with copyright and privacy laws.
Patent landscape: Review patents to avoid infringement in AI software.
Contracts: Clearly define ownership of derived AI models and outputs.
5. Practical Steps for IP Due Diligence in AI
Inventory all IP assets (code, datasets, models, algorithms)
Verify ownership and assign IP rights correctly
Conduct patent and copyright searches relevant to AI
Audit employee access and exit policies
Check third-party software and open-source licenses
Review cloud service agreements and security measures
Assess regulatory compliance affecting datasets
RELATED Blog
comments