Hackers Exploit Prompt Engineering to Manipulate AI Agents, Threatening Business Security

Hackers are using prompt engineering to manipulate AI agents, posing a threat to business security. Gartner predicts that by 2028, 15% of daily decisions will be made autonomously by AI. However, the massive integration and excessive permissions given to AI make it a new target for cybercrime. Techniques such as steganographic prompting are being misused, leading to the execution of unauthorized operations or the disclosure of confidential information. The issue is urgent and serious.