Cardano founder Charles Hoskinson has raised concerns about the implications of artificial intelligence for intellectual property, arguing that researchers could unknowingly expose unpublished work when using cloud-based AI systems.
Hoskinson discussed the issue in a recent YouTube video while commenting on a post by mathematician Jay Cummings, a professor at the University of California, San Diego.
Cummings had shared an OpenAI post describing a mathematical breakthrough while alleging that his own work on the problem had been used by ChatGPT. According to Hoskinson, Cummings also had chat logs that could potentially support his account.
Hoskinson questioned the suggestion that the AI system had arrived at the result entirely independently.
“OpenAI says that they totally independently came up with this,” Hoskinson said, while arguing that the available evidence could warrant closer examination.
He suggested that when an AI system is confronted with an unusually difficult mathematical problem, it may search through information available to it rather than necessarily developing a completely independent approach.
Hoskinson characterized AI models as optimization systems that attempt to find an efficient route toward a solution. In the scenario he described, material previously provided through ChatGPT conversations could potentially become relevant to a model’s response.
The Cardano founder extended the argument beyond mathematics, warning that academics, researchers and entrepreneurs should consider the implications of sharing unpublished ideas with cloud-based AI services.
“If you’re an academic, if you’re an entrepreneur, know that your ideas, if you share them in AI with these frontier models in the cloud, they’re not your ideas anymore,” Hoskinson said.
Hoskinson argued that producing a solution to a longstanding mathematical problem can represent a major professional achievement, making the protection of the underlying research especially important.
Midnight’s solution
Hoskinson subsequently connected the issue to Midnight, a privacy-focused blockchain project associated with the Cardano ecosystem.
He argued that AI systems running in privacy-preserving environments could allow users to access inference capabilities without giving infrastructure operators access to their conversations or other sensitive information.
“Why does Midnight fix this?” Hoskinson said. “Because it creates private environments that you can run the AI in. So they can’t read your logs.”
Midnight is being developed around a distinct economic model intended to separate network security and governance from the resources required to use the network. The architecture is designed to address some of the volatility associated with using a single cryptocurrency for both network participation and transaction costs.
Under this model, users could potentially interact with applications without having to directly manage a volatile native asset for every transaction.
“Satoshi gave us good money; Ethereum gave us programmability; Cardano brought the third generation of interoperability, scale and good governance. Midnight gives us our identity and privacy back,” Hoskinson wrote.
Hoskinson previously discussed a broader shift toward protocol-level revenue and economic sustainability for Midnight, arguing that network utility could eventually play a greater role in supporting the ecosystem than external incentives.