May 28th – Earlier this month, OpenAI launched ChatGPT-4o, a more advanced version of the 4.0 model. This latest model can process audio, visuals and text in real-time. It’s available to all users for free, although the service does have some limitations.
In relation to cyber security, experts advise cyber security professionals to temper expectations surrounding any profound impact on the cyber security landscape.
The model is not expected to significantly alter how generative AI influences either cyber attackers or cyber defenders.
According to machine research scientist Daniel Kang, cyber adversaries may be able to use the model for multi-tasking purposes, thereby seeing minor productivity boosts.
Voice cloning concerns
One of ChatGPT-4o’s new features includes voice cloning, a process that cyber criminals may weaponize in order to commit fraud.
For example, a cyber criminal could identify a target, obtain a voice recording of them via YouTube, Facebook or TikTok, clone the voice via ChatGPT-4o, and use it to break into a bank account.
Voice cloning capabilities are also expected to exacerbate issues around misinformation.
ChatGPT-4o security
Cyber security around the ChatGPT-4o model remains in question. OpenAI says that it’s added security and privacy safeguards. In particular, OpenAI has reduced the amount of data that it collects, effectively anonymizes that data, and uses stronger encryption protocols than previously.
CISOs are advised to continue developing and implementing generative-AI related policies and safeguards. Cyber security teams also need to consider how to manage third-party risk.
For further generative AI insights from Cyber Talk, please click here. Lastly, to receive stellar cyber insights, groundbreaking research and emerging threat analyses each week, subscribe to the CyberTalk.org newsletter.