GitHub Copilot, Copyright Infringement and Open Source Licensing

Users of GitHub Copilot can produce source code subject to license requirements without attributing and licensing the code to the rights holder.

Risk classification

• Primary risk domain: 6 Socioeconomic & Environmental
• Primary risk subdomain: 6.3 Economic and cultural devaluation of human effort

The incident involves an AI system reproducing human creative work (software code) without proper attribution, potentially devaluing the effort of open-source developers and violating licensing terms.

Additional risk subdomains

• 7.3 Lack of capability or robustness: The AI system occasionally regurgitates verbatim training data, demonstrating a lack of robustness in preventing copyright and licensing violations.

Causal factors

• Entity: AI
• Intent: Unintentional
• Timing: Post-deployment

The risk of license non-compliance arises from the AI system's post-deployment generation of verbatim code snippets, which was an unintentional outcome of its training process.

EU AI Act risk tier

• Risk tier: 4 Minimal or No Risk

Minimal or No Risk: The system is an AI-powered code completion tool used for software development, which poses low or negligible risks to safety or fundamental rights under the EU AI Act framework.

AI system and alleged parties

• AI system: GitHub Copilot, OpenAI Codex (GitHub, OpenAI)
• AI purpose: Code Generation; Writing Assistant
• Behaviour type: Assistant
• Alleged developer: GitHub
• Alleged deployer: programmers, GitHub
• Alleged harmed parties: Intellectual Property rights holders

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Severe. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Negligible
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Negligible, indirect Negligible
• Differential treatment: direct Negligible, indirect Negligible
• Civil rights: direct Negligible, indirect Negligible
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Severe, indirect Negligible
• Psychological: direct Negligible, indirect Negligible
• Epistemic: direct Negligible, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Privacy

Reported: Yes, the lawsuit explicitly mentions violations of GitHub's own privacy policy.

Directly caused: The lawsuit contends that the defendants violated GitHub's privacy policy by training their AI systems on public repositories without proper consent or compliance.

Inferred additional harm: It is likely that personal data or sensitive information contained within public repositories was processed and memorized by the model without explicit user consent.

People affected

• Occurrences reported: 1
• People reportedly harmed: 1000000
• People reportedly exposed: 1000000

Potential causes

Management

• Prioritizing Speed Over Compliance: Launching technical preview before resolving licensing and legal questions.
• Reliance on Broad Terms of Service: Using GitHub's ToS to justify training on hosted code without extra consent.

Technology

• Codex Verbatim Regurgitation: The underlying model sometimes reproduces exact snippets of training code.
• Lack of Similarity Filter: Technical preview lacked tools to check if output matched licensed code.

Data Inputs

• Use of Copyleft Licensed Code: Training data included GPL and MIT code requiring attribution.
• Training on External Code: Using code hosted outside GitHub without explicit terms of service consent.

Human Factors

• Blind Trust in Suggestions: Developers accepting generated code without verifying license compliance.
• Overreliance on Fair Use Assumptions: Assuming machine learning training is universally protected under fair use.

Process and Methods

• No Output Attribution Process: The tool did not automatically attribute original authors of copied code.
• Inadequate Code Review: No robust process to filter out copyrighted comments and code blocks.

Regulatory Environment

• Unsettled Fair Use Precedent: No definitive court rulings on AI training on copyrighted code.
• Ambiguous Derivative Work Definitions: Legal uncertainty on whether AI-generated code constitutes a derivative work.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide clear, detailed accounts of the launch of GitHub Copilot, the legal arguments surrounding fair use and copyright, and the filing of the class-action lawsuit. The facts of the case are well-documented across multiple sources.
• Ambiguities identified: The exact frequency of verbatim code regurgitation is disputed, with GitHub claiming 0.1% and critics showing specific examples of longer blocks with comments.
• Alternative interpretations: Some might interpret the incident primarily as a contract/licensing dispute rather than an AI safety failure, though the underlying cause is the AI's training methodology and output behavior.

The GitHub Copilot licensing controversy and subsequent class-action lawsuit highlight emerging legal and intellectual property challenges in the generative AI era. While it poses negligible direct threat to physical safety or national sovereignty, it establishes important legal precedents for economic and technological security frameworks regarding how AI models utilize public data.

• Overall national security impact: Minor
• Scope: Multiple nations
• Primary target: No clear primary
• Other affected: Global open-source software community
• Alleged perpetrator: Unknown

Threat characteristics

• Imminence: Long-term. The incident represents an ongoing legal and strategic debate regarding AI training practices rather than an active, imminent crisis.
• Autonomy: Human-controlled. The AI serves as an assistant suggesting code, leaving the final decision to accept, reject, or modify the code entirely to the human developer.
• Novelty: First-of-its-kind. This represents one of the first major class-action lawsuits tackling the novel legal challenges of generative AI training on copyrighted data.

Impact by dimension

• Physical security: Negligible. The incident is a software licensing and copyright dispute with no impact on physical systems, critical infrastructure, or human safety.
• Information security: Negligible. No intelligence compromise, classified data theft, or systematic information warfare operations are associated with this incident.
• Sovereignty: Negligible. This civil legal dispute does not threaten state authority, territorial control, or core government operations.
• Economic security: Minor. The dispute over intellectual property and code regurgitation impacts the legal and economic frameworks governing software development, which indirectly affects technological competitiveness.
• Societal stability: Negligible. The incident involves digital rights and licensing issues, presenting no threat to social cohesion, civil liberties, or public safety.