🔪 AI agents – a disaster for the crypto industry…

Warning from CertiK CEO: a mass deployment of AI agents without isolation and verification could become a catastrophe for cybersecurity.

➥ AI agents are already gaining access to files, emails, databases, local keys, wallets, and financial services. If such an agent is not isolated, it becomes an internal threat: it can be easily hacked not with a virus, but with plain text.

➥ Through "prompt injection," a hacker can hide malicious instructions in a PDF, email, or website. The agent will read it as part of its task, change its rules, and be able to leak data or initiate unauthorized transactions.

➥ CertiK has already found hundreds of malicious plugins, fake installers, and counterfeit dependencies for AI agents. Traditional antivirus programs hardly detect them because the attack does not come through typical malicious code.

➥ CertiK reports a surge in short-lived on-chain traps: scammers launch schemes lasting 10 minutes or a few hours to deceive not people, but AI bots and trading agents.

➥ According to CertiK's CEO, the crypto industry must urgently adopt "Zero Trust" for AI infrastructure: every tool, team, and connection must be verified separately.