May 9, 2024 at 11:18AM
Trend Micro’s 2023 investigation at the 2024 RSA Conference revealed criminals continue to lag in AI adoption. They identified a criminal LLM called WormGPT and potential scams, like EscapeGPT and FraudGPT. Criminals prefer mainstream AI products over building their own systems and use deepfake services for illicit activities. Trend Micro foresees an increase in criminal use of AI.
From the meeting notes, there are a few key takeaways:
1. Criminals are still lagging behind in the adoption of AI.
2. There is a growing incidence of jailbreaking services and other unclear “services” being offered, many of which are being labeled as potential “scams”.
3. Criminals are currently focusing on the use of mainstream AI products and incorporating AI into their existing methodologies, rather than developing their own AI systems.
4. The use of deepfake services is increasing, making it easier to create realistic deepfakes for broad attacks.
5. Jailbreaking services allow criminals to use existing language model models with minimal risk of being tracked and traced, providing an anonymized access to them.
6. Criminals are cautious and methodical in including the capabilities of AI into their activities.
7. Defenders currently have the edge in the AI cat and mouse game between criminals and defenders.
These takeaways provide insights into the current landscape of criminal use of AI and the potential future developments in this space.