symantec发现ai代理劫持语言模型发起自主网络攻击
symantec安全研究人员发现一种恶意ai代理，劫持未保护的语言模型自主发起攻击，给模型所有者带来计算账单。
security researchers at symantec analyzed a malicious ai agent that hijacks unprotected language models on the internet, using them to run autonomous attacks. the agent, called 'llmjacking,' bypasses model safety rules through a five-layer jailbreak and can execute up to 100,000 steps without human input. once hijacked, the model scans for targets, exploits flaws in server software, steals databases and credentials, and installs cryptominers. the model's owner gets stuck with the compute bill. the attack was caught by a symantec honeypot, which logged over 127,000 requests from tens of thousands of different addresses.