How AI is opening new doors for hackers to cause chaos

AI is providing new opportunities for hackers, as they exploit open-source models to infiltrate companies using traditional methods. These models, while popular for their accessibility and cost-effectiveness, pose significant cybersecurity risks due to the potential for malicious code to be embedded within them. JFrog's CTO, Yuval Fernbach, highlights the increasing prevalence of attacks, particularly those that overload models to render them unresponsive. Despite efforts by industry leaders to mitigate these threats, the responsibility largely falls on individual companies to ensure their models are secure. A survey by JFrog and InformationWeek found that many companies lack policies or mechanisms to review the use of open-source AI models, increasing their vulnerability.

The use of open-source models is rising as companies seek to avoid the costs and data-sharing concerns associated with proprietary models. However, this trend is accompanied by a growing number of attacks, with JFrog's research revealing that a small percentage of models on platforms like Hugging Face contain malicious code. The rapid advancement of AI, particularly generative AI, is outpacing the ability of industries like banking to respond effectively, leading to costly errors. As AI continues to evolve, the integration of multiple models with varying capabilities will necessitate more rigorous security checks to prevent malicious activity and ensure reliable outcomes.

Key takeaways

Hackers are using AI models to infiltrate companies, with open-source models raising cybersecurity concerns.
Malicious code can be hidden in open-source large language models, leading to increased attacks.
Many companies lack policies or methods to review developers' use of open-source AI models.
The rise of agentic AI increases risks as models perform tasks beyond providing information and analysis.

How AI is opening new doors for hackers to cause chaos

Key takeaways

Discussion (0)