커넥티비티 클라우드 기업인 클라우드플레어(Cloudflare)가 5일 거대언어모델(LLM)을 상대로 한 남용과 공격을 식별하는 새로운 보호 레이어(layer)인 ‘AI용 방화벽(Firewall for AI)’을 선보인다고 밝혔다.
Firewall for AI, rapid risk detection and automatic blocking
AI model can respond to cyber threats targeting companies
Cloudflare supports a firewall that can protect against cybersecurity threats for businesses that utilize AI models.
Cloudflare, a connectivity cloud company, announced on the 5th that it is introducing 'Firewall for AI', a new protection layer that identifies abuse and attacks against large language models (LLMs).
Cloudflare aims to help customers respond to attacks targeting functions, critical data, and trade secrets within LLMs through its AI-powered firewall, built on the world’s largest global network.
According to data from Cloudflare, only one in four C-suite executives are confident their organization is prepared for the risks posed by AI. LLMs are often left defenseless against threat actors because it is impossible to restrict user interaction and they are difficult to pre-design.
According to market research firm Gartner, generative AI applications cannot be secured in isolation, and a solid foundation of cloud, data, and application security must be in place before planning and deploying generative AI-specific security controls.
Cloudflare’s ‘Firewall for AI’ adds a layer to its existing comprehensive security platform to ultimately block threats from emerging technologies. “With Firewall for AI, LLM applications can be protected from potential vulnerabilities that could be used as weaponization of AI models,” Cloudflare said.
Firewall for AI can be deployed in front of any LLM running on Cloudflare’s Workers AI. It scans and evaluates user-submitted prompts to better identify attempts to exploit the model and extract data, enabling faster threat detection. All customers running LLM on Cloudflare’s Workers AI are protected by Firewall for AI for free, helping to prevent concerns such as prompt injection and data exfiltration.
It can also automatically block threats without human intervention. Built on Cloudflare’s global network, the AI-powered firewall is deployed close to end users, providing unprecedented protection from model abuse almost immediately.
“As new types of applications emerge, new types of threats quickly follow,” said Matthew Prince, CEO and co-founder of Cloudflare. “The same goes for AI-based applications,” he said. “Cloudflare is helping to build security into AI environments from the ground up with our AI firewall. We want to provide one of the first protective shields for AI models, so that enterprises can capitalize on the opportunities presented by technological advancements while staying protected.”