The cybersecurity company behind the exposure of WormGPT has published a blog post. This provides information about the strategies used by cybercriminals who “jailbreak” popular AI chatbots like ChatGPT. This refers to tactics that circumvent the security limits that companies impose on their chatbots.
SlashNext researchers have found that cybercriminals don't just share their successful jailbreaks on discussion forums to make them accessible to others. Instead, developers are also promoting AI bots that act as criminals
purposes can be used. They claim these are custom language models (LLMs). SlashNext has confirmed that this is not true in most cases, but rather that it concerns jailbroken versions of public chatbots like
ChatGPT works. These include tools such as EscapeGPT, BadGPT, DarkGPT and Black Hat GPT.
The advantage for cybercriminals who use one of these tools instead of jailbreaking ChatGPT themselves is that their identities remain completely anonymous. “Jailbreaking” and the use of generative AI to increase phishing effectiveness are currently enjoying an interesting hype. However, there is still little evidence that they really represent a significant innovation. While there are certainly advantages for non-native speakers to create better phishing texts or for novice programmers who can hack together malware more quickly, there is nothing to suggest that professional cybercriminals are taking advantage of AI. Sellers benefit from buyers not doing enough research and falling for offers that sound attractive.
Unclear threat situation
When it came to the topic of “jailbroken” LLMs, my first thought was that malicious actors could compromise the AI-driven chatbots that are ubiquitous on legitimate websites. To me this would pose a greater danger to normal
Present to consumers as a phishing email with improved grammar. That's not to say that GPT-style AI isn't a threat. In fact, we have not yet figured out what exactly this threat is. Through the increased attention
The future of AI in cybersecurity will be closely examined. Hopefully, the more serious vulnerabilities can be closed before they are ever exploited. (Chris Vaughan, VP Technical Account Management at Tanium)
About Tanium Tanium, the industry's only Converged Endpoint Management (XEM) provider, is leading the paradigm shift in traditional approaches to managing complex security and technology environments. Only Tanium protects every team, endpoint, and workflow from cyber threats by integrating IT, compliance, security, and risk into a single platform. The Tanium platform provides comprehensive visibility across all devices, a unified set of controls, and a common taxonomy.