The world of artificial intelligence is going through a period where the line between a revolutionary productivity tool and a sophisticated weapon in the hands of state actors is constantly blurring. Recent reports from OpenAI confirm that generative models like ChatGPT have become targets of targeted attacks by Chinese and Iranian hacking groups. These groups do not use AI for simple text writing, but for systematically spreading malware and managing complex influence campaigns.

Cybernetic Weapon in the Digital Age: Malware and Phishing

According to security expert analyses, hackers are focusing on using LLMs (large language models) to overcome traditional barriers in creating malicious code. Although modern models have built-in security filters, attackers are looking for ways to bypass these mechanisms using so-called jailbreaking or sophisticated prompting.

Among the most significant identified incidents are:

• SweetSpecter: A Chinese operation targeting Asian government entities, which used spear-phishing (targeted phishing) to gain access to sensitive data.
• CyberAv3ngers: An Iranian group that attempted to use ChatGPT to extract passwords from the macOS operating system.
• Storm-0817: An attack targeting Android devices, where AI helped generate convincing communication templates for stealing contacts and browsing history.

From a technical perspective, this is a dangerous shift. While previously a hacker had to write malicious code manually or using specialized tools, now ChatGPT can serve as an "intelligent assistant" that helps with the logical structure of an attack or creates text that, in the context of a phishing email, does not appear to be an automatic translation, but an authentic message from a trusted person.

Disinformation Campaigns: "Data Center Bandwagon" and Political Manipulation

In addition to direct technical attacks, AI is becoming a key tool for hybrid warfare. OpenAI identified two coordinated campaigns codenamed "Data Center Bandwagon" and "Tech and Tariffs". These operations aimed to deepen political divisions in the USA by massively generating comments, social media posts, and even political cartoons.

Chinese operators exploited real social issues – for example, community resistance to the construction of new data centers. Using ChatGPT, they were able to generate a huge amount of content that looked like organic discussions by real people, but in reality, it was a controlled disinformation operation. Although analyses show that these campaigns did not gain massive traction, it demonstrates the extreme efficiency with which AI can be used to "flood" the internet with fake content.

Security Comparison: OpenAI vs. Competition