When does nondisclosure of AI work, and when does it backfire?
The key factor is whether your customers can tell you're using AI. A 2025 game-theoretic study found that nondisclosure is ineffective in markets with a high proportion of 'sophisticated' consumers — people who can recognize AI involvement [1]. In those markets, trying to hide AI use actually hurts you. But in markets dominated by 'naive' consumers who can't easily detect AI, nondisclosure remains a viable strategy, especially when the AI service quality is perceived as low (to avoid backlash) [1]. This means sovereignty strategies that involve withholding information about AI use can work, but only if your audience isn't savvy enough to spot it.
The same study also found that consumer aversion to AI doesn't automatically push merchants toward nondisclosure. Instead, the effectiveness of hiding AI depends more on service quality than on how much people dislike AI [1]. So if you're providing a genuinely good AI service, you might not need to disclose it — but if your service is poor, hiding it won't save you.
Can data sovereignty strategies actually reduce barriers in AI pipelines?
Yes, and there's real-world evidence to back this up. A 2022 action research project at Mondragon Corporation (a large industrial group) integrated a data sovereignty component into an existing AI pipeline where sensor data was collected and sent to a data quality service provider [3]. Over 12 months, the researchers identified 10 lessons learned, 4 benefits, and 3 barriers to data-sovereign AI pipelines. Their key finding: a data sovereignty component can help reduce existing barriers and increase the success of collaborative data science initiatives — even when providers disclose less data [3]. This suggests that sovereignty strategies aren't just theoretical; they can work in practice to make data sharing easier and more trustworthy.
However, the same study also identified barriers, meaning sovereignty solutions aren't a magic bullet. They require careful implementation and may introduce their own complexities [3]. So while they can work, you need to plan for the challenges.
Is mandatory disclosure always the best policy?
Surprisingly, no. The 2025 game-theoretic study found that mandating AI identity transparency does not always maximize social welfare, challenging the assumption that compulsory disclosure is universally beneficial [1]. This is a crucial nuance: forcing providers to disclose everything about their AI systems might actually reduce overall welfare in some situations. For example, if disclosure leads to consumer backlash against a genuinely useful AI service, everyone loses. This means that sovereignty strategies that involve selective disclosure — or even nondisclosure — can sometimes be better for society as a whole, not just for the provider.
This finding is reinforced by a 2023 paper on digital sovereignty, which notes that both authoritarian and democratic states use the concept of digital sovereignty to justify very different policies [4]. There's no one-size-fits-all approach. The same paper argues that the HCI (human-computer interaction) community needs to develop human-centered definitions of digital sovereignty that strengthen the position of users under non-sovereign conditions [4]. This suggests that effective sovereignty strategies must be tailored to the specific context and power dynamics, not based on a blanket rule about disclosure.
About These Sources
This answer is built on 5 peer-reviewed studies — published from 2022 to 2025, 2 from 2024 or later, 1 in Q1 journals — selected as the most relevant from 5 studies that passed quality screening, drawn from 44 papers retrieved from a database of over 500 million.
Sources used in this answer
To reveal or conceal: AI identity disclosure strategies for merchants
A game-theoretic model found that nondisclosure of AI identity is ineffective in markets with many sophisticated consumers who can detect AI, but works in markets dominated by naive consumers, especially when AI service quality is low; mandating transparency does not always maximize social welfare [1].
Indigenous Data Sovereignty: A Catalyst for Ethical AI in Business
A commentary argues that Indigenous data sovereignty is a powerful framework for resisting digital colonialism and promoting ethical AI development, though it does not provide empirical data on disclosure strategies [2].
Data sovereignty for AI pipelines
A 12-month action research project at Mondragon Corporation found that integrating a data sovereignty component into an AI pipeline reduced barriers and increased success in collaborative data science, identifying 10 lessons learned, 4 benefits, and 3 barriers [3].
Digital Sovereignty: What it is and why it matters for HCI
A conceptual paper argues that digital sovereignty is used differently by authoritarian and democratic states, and calls for human-centered definitions to strengthen user positions under non-sovereign conditions [4].
Data Disclosure in AI Invention
A legal analysis (in Korean) discusses the challenge of technology disclosure in AI inventions due to the 'black box' nature of neural networks, drawing parallels to microbial inventions and arguing for adapted disclosure requirements in patent law [5].