In an effort to boost the security of artificial intelligence applications, Cloudflare has unveiled its Firewall for AI – a new layer of protection designed to reduce and prevent attacks that target large language models (LLM).
LLMs play an important role in interpreting human language and other complex data, and can be trained on billions of parameters, however just like any other digital asset, threat actors have been seeking to exploit vulnerabilities.
With AI now at full pace, Cloudflare says its Firewall for AI comes at a time when proactive security measures are quickly becoming critical.
Cloudflare wants to secure AI at the LLM level
The company’s announcement cites a recent study revealing that only one in four C-Suite execs felt confident in their organization’s readiness to address the risks of artificial intelligence.
However, in addressing their concern, Cloudflare said that securing LLMs poses its own challenges because it can be difficult to apply sufficient protection without hindering user interactions, despite their susceptibility to tampering, attacks and abuse.
Cloudflare CEO Matthew Prince emphasized the important nature of addressing emerging AI threats: “When new types of applications emerge, new types of threats follow quickly. That's no different for AI-powered applications.”
The company’s new Firewall for AI is set to equip security teams with new tools to protect their LLM applications effectively, including rapid threat detection and automatic blocking of threats without the need for human intervention.
Despite the promising advancement, it appears that Cloudflare thinks this is just one pillar in a multifaceted approach, citing a Gartner statement: “You cannot secure a GenAI application in isolation. Always start with a solid foundation of cloud security, data security and application security, before planning and deploying GenAI-specific security controls.”
The service is set to be available to customers using LLMs on Cloudflare’s Worker AI platform free of charge.
More from TechRadar Pro
- Check out the best AI tools and the best AI writers
- We’ve rounded up a list of the best cloud hosting providers
- AI models could be attacked, flawed by this Hugging Face security issue — security worries add to AI concerns
