Vitalik:兼顾安全、去中心化与隐私的AI代理设计仍面临挑战

PANews 3月11日消息,以太坊创始人Vitalik Buterin在X平台发文称,在保证安全、去中心化和隐私的前提下,应对AI代理可能面临的对抗性攻击仍存在困难。例如,你的AI代理查看了对方的ENS资料,而这个资料里藏着一个越狱指令,导致代理把你所有的币都转给了对方。
Vitalik认为,让每一笔大额交易都需要人工确认,比完全没有确认要好得多,但这仍然不完美。他还提出,应该向用户解释这笔交易的具体内容,这是一个复杂的用户体验问题。
Disclaimer: OKX Orbit content is provided for informational purposes only. Learn more
Replies
Related Flash News
Three major DeFi attacks in five days, with a total of 14 crypto hacking incidents in May
Kelp: rsETH recovery makes key progress, with multiple DeFi protocols collectively liquidating attacker positions
Echo Protocol confirmed the attack and suspended all cross-chain transactions
Data: A smart money bought 1,344.18 ETH on the chain again 10 hours ago, and now has a floating loss of $165,000
Echo Protocol suffered an attack on the Monad chain, where hackers minted 1,000 eBTC and then withdrew funds through Curvance
Lido: Swellchain will be suspended on June 15, and users need to withdraw wstETH assets in advance
Data: ETH lending protocol TVL has dropped from a year-to-year high of $32 billion to $23 billion
Two researchers at the Ethereum Foundation announced their resignation
Aster launches on-chain listing proposal voting mechanism
Delphi Digital: Aave's three core markets have structural problems, with an annualized loss of about $52 million



