CoinDesk
CertiK co-founder and CEO Ronghui Gu is sounding the alarm on the global rush to deploy autonomous AI agents, warning that the industry is accumulating a catastrophic "security debt" by pushing unisolated, unvetted systems into production at scale. Speaking to CoinDesk following CertiK's deep-dive report into agent infrastructure, Gu said the moment a user grants an AI agent access to local files, credentials, email, or financial tools, that agent becomes "the ultimate inside threat" — one that can be silently hijacked without a single line of malicious code.
The attack vector is deceptively simple: prompt-injection attacks embed hidden instructions inside ordinary documents, webpages, or emails.
CoinTelegraph
WuBlockchain
Crypto Capital Venture
CryptoSlate