The recent settlement between Anthropic and a group of authors was meant to resolve a long-running dispute over the use of copyrighted books in AI training data. But now a new conflict has emerged: authors are pushing back as publishers and literary agents seek a share of the settlement proceeds. This development, reported by TechCrunch, underscores the unresolved complexities of AI training data rights—complexities that extend well beyond the publishing industry and into sectors like crypto, where AI is increasingly integrated into operations.
For crypto projects, the Anthropic case is a cautionary tale. Many blockchain-based ventures use AI for tasks such as fraud detection, market analysis, and content generation. If those AI models are trained on copyrighted material without proper licensing, the projects could face legal liabilities similar to those Anthropic encountered. The authors' pushback highlights that even when a settlement is reached, the distribution of compensation can be contentious, and the underlying legal questions remain unsettled.
Why It Matters
The Anthropic settlement is not just a publishing industry story. It sets a precedent for how AI companies—and by extension, any organisation using AI—handle copyrighted training data. For crypto projects, which often operate in a regulatory grey area, the risk is amplified. A lawsuit over training data could not only result in financial penalties but also damage investor confidence and disrupt operations.
Moreover, the dispute over who gets paid from the settlement reveals a broader issue: the lack of clear legal frameworks for AI training data. Until courts or regulators provide definitive guidance, businesses using AI must proceed with caution, ensuring they have proper licences or using datasets that are clearly in the public domain.
The Settlement and the Pushback
According to TechCrunch, Anthropic reached a settlement with a group of authors who had sued over the use of their books in training AI models. The settlement was intended to compensate the authors, but now publishers and agents are claiming a portion of the funds, arguing that they hold rights to the works. Authors are pushing back, asserting that the compensation should go directly to them.
This internal conflict is significant because it shows that even when a tech company agrees to pay for training data, the question of who is entitled to that payment is not straightforward. For crypto projects, this means that simply obtaining a licence from one party may not be sufficient; they may need to ensure they have rights from all relevant stakeholders, including authors, publishers, and agents.
Implications for Crypto Projects Using AI
Crypto projects often rely on AI for various functions, from algorithmic trading to NFT content generation. If these projects use AI models trained on copyrighted data without proper authorisation, they could face legal action. The Anthropic case demonstrates that copyright holders are willing to sue, and that settlements can be costly.
Moreover, the dispute over settlement distribution suggests that even if a project reaches a settlement, it may not be the end of the matter. Other parties may come forward with claims, leading to prolonged legal battles. This uncertainty can be a significant operational risk for crypto startups, which often have limited legal resources.
To mitigate these risks, crypto projects should:
- Conduct thorough due diligence on AI training data sources.
- Use AI models that are trained on licensed or open-source data.
- Implement clear policies for AI-generated content to avoid copyright infringement.
- Stay informed about legal developments in AI and copyright.
Commercial Impact
The commercial impact of this dispute is twofold. First, for AI companies like Anthropic, the settlement and subsequent disputes could increase the cost of using copyrighted training data. This cost may be passed on to customers, including crypto projects that use AI services. Second, for crypto projects that develop their own AI models, the legal uncertainty could deter investment and innovation.
On the other hand, there is a potential opportunity for crypto projects that can demonstrate compliance with copyright laws. They could use this as a competitive advantage, attracting investors and customers who are concerned about legal risks.
Risks and Unknowns
The main risk is that the legal landscape for AI training data remains unclear. The Anthropic settlement does not set a binding precedent, and other cases are still pending. Additionally, the authors' pushback could lead to further litigation, which might result in different outcomes.
Another unknown is how regulators will respond. In the UK, for example, the government has been considering changes to copyright law to accommodate AI. According to a BBC report, the Chancellor has said the UK economy is 'turning a corner' despite debt concerns, but the report does not address AI copyright specifically. However, any regulatory changes could have significant implications for crypto projects using AI.
FY Outlook
In the near term, we expect continued legal battles over AI training data. The Anthropic settlement is unlikely to be the last, and the authors' pushback may encourage other groups to challenge similar agreements. For crypto projects, the prudent approach is to assume that using copyrighted training data without clear rights is risky.
Longer term, we anticipate that courts and regulators will eventually provide more clarity. In the meantime, crypto projects should focus on building AI systems that are legally sound, perhaps by using synthetic data or data that is explicitly licensed for AI training.
Conclusion
The authors' pushback on the Anthropic settlement is a reminder that AI training data is a legal minefield. For crypto projects, the implications are clear: they must be proactive in managing copyright risks to avoid costly litigation and reputational damage. By staying informed and adopting best practices, they can navigate this uncertain terrain.
Sources and References
The reporting and evidence for this briefing were checked against techcrunch.com (techcrunch.com) and bbc.co.uk (bbc.co.uk).
