IC3 said in an X report on June 8 that machine learning models can significantly improve smart contract security and fraud detection; AI-driven trading systems may enable collusion between autonomous agents and create unfair advantages; crypto infrastructure can build tamper-resistant data pipelines for AI model training; currently, there is no publicly available quantitative evidence proving that decentralized AI pipelines can effectively reduce end-to-end costs or improve performance metrics.
Four core conclusions of the report
Four research findings confirmed in the IC3 report:
AI makes crypto more flexible: Machine learning models can significantly improve smart contract security, strengthen real-world data processing capabilities, and optimize fraud detection
New avenue for market abuse: AI-driven trading systems may allow autonomous agents to collude with each other and create unfair internal advantages through opaque strategies
Cryptography secures the AI supply chain: Crypto infrastructure can establish highly secure, trustworthy, and tamper-resistant data pipelines for AI model training
Decentralized real-world validation: At present, there is almost no public quantitative evidence to show that decentralized AI pipelines can actually reduce end-to-end costs or improve metrics
Ari Juels’ key technical observations
In the report, Ari Juels pointed out the fundamental difference between two technologies: cryptography is a “hard” technology, built on cryptographic primitives and clearly defined procedures with strict security properties; AI is a “soft” technology—no one can fully understand or fully trust the models it relies on. He said that simply combining the two is “like welding jelly”; but if combined properly, cryptography can convert AI’s liquidity into systems that are secure, reliable, and highly autonomous.
Giulia Fanti also noted that the sheer volume of research makes it very difficult to distinguish useful information from useless information. The report aims to outline future blockchain research directions for academia over the next ten years, and to provide corporate leaders with an R&D roadmap.
FAQ
What specific evidence does the IC3 report use to evaluate decentralized AI?
The IC3 report explains that although the industry heavily promotes the advantages of decentralized AI pipelines, there is currently almost no publicly available, quantitative evidence proving that decentralized AI pipelines can actually reduce end-to-end costs or improve effectiveness metrics. The report does not entirely deny the potential of decentralized AI, but instead points to a lack of publicly verified data.
What exactly does “cryptography secures the AI supply chain” mean?
According to the IC3 report, crypto infrastructure can build highly secure, trustworthy, and tamper-resistant data pipelines for training AI models. The significance of this direction is to ensure that the data sources used to train AI models are trustworthy and not maliciously tampered with, thereby improving the overall trustworthiness of AI systems.
Which readers will find this report most valuable?
At the time of the report’s release, Giulia Fanti explained that the report maps out development directions for blockchain research in the next ten years for academia, while also providing an important R&D roadmap for enterprise leaders. The report was written jointly by more than 20 researchers from both industry and academia, and took several months.
