Telegram's Pavel Durov Unveils Decentralized AI Network on TON Blockchain

According to Cointelegraph, Pavel Durov, co-founder of Telegram, has announced the development of a decentralized AI network called the Confidential Compute Open Network (Cocoon). This initiative will be built on The Open Network (TON), a layer-1 blockchain linked to Telegram. Durov made this announcement at the Blockchain Life 2025 forum in Abu Dhabi, emphasizing the network's aim to provide AI-driven features while maintaining user data privacy, unlike centralized AI providers. Durov explained that users could contribute their graphics processing units (GPUs) to the network and, in return, receive Toncoin (TON), the native token of TON. He highlighted the importance of decentralized AI for preserving human freedom, noting that the world has been losing digital freedoms over the past two decades. This move towards decentralization is seen as crucial in countering the trend of centralization, which often compromises user privacy and freedom. The topic of decentralizing AI models has gained traction among AI and blockchain developers due to privacy concerns and the potential for centralized services to censor or manipulate information without user awareness. Centralized AI systems are vulnerable to data breaches and hacks, as they store large amounts of user data on centralized servers, making them attractive targets for cybercriminals. David Holtzman, chief strategy officer of the Naoris decentralized security protocol, highlighted these risks in discussions with Cointelegraph. Moreover, centralized AI providers have the ability to alter algorithms or distort data in real-time, potentially influencing public opinion. Blockchain technology offers a solution by ensuring that AI-generated data remains tamper-proof. By using a decentralized ledger, blockchain can record the origin and chain of custody of data, creating an immutable and verifiable digital record onchain. This capability underscores the potential of blockchain to enhance data integrity and security in AI applications.