• A recent Deloitte study found that just 25% of C-suite executives are sure that their companies are ready to handle the threats that AI presents.
  • According to the business, the new Firewall for AI will enable security teams to quickly identify new threats because it can be installed in front of any LLM that is using Cloudflare’s Workers AI product.

Global provider of cloud connectivity, Cloudflare Inc., announced a Firewall for AI, a layer of defense for artificial intelligence large language models that aims to detect possible attacks before they can access sensitive data or tamper with crucial functionality.

The business also unveiled a new line of defensive cybersecurity technologies that employ AI to counter new attacks. These consist of identifying unusual user activity, filtering email to identify questionable content, and reducing risks to the company.

As an increasing number of organizations adopt LLMs and AI models as integral components of digital transformation, the enterprise is confronted with the associated security risks. A recent study by Deloitte reveals that a mere one-fourth of C-suite-level administrators possess confidence in the readiness of their organizations to confront the risks presented by artificial intelligence.

Matthew Prince, Co-founder and Chief Executive of Cloudflare, said, “We’ve entered into the AI arms race, and it’s critical that today’s AI-powered apps — many of which fuel our healthcare, our banking systems, and our grid — are built on secure models. This protection should be available to all because a secure internet is better for everyone.”

According to the business, the new Firewall for AI will enable security teams to quickly identify new threats because it can be installed in front of any LLM that is using Cloudflare’s Workers AI product. By placing any LLM as close to enterprise consumers as possible, Workers AI enables developers to deploy AI models at the edge at scale over Cloudflare’s global network, enabling exceptionally low latency responses.

The company claims that by placing the firewall in front of the LLM, it is possible to identify attempts to exploit the model and extract data by scanning user-submitted queries. Consequently, it is capable of blocking hazards autonomously and without human intervention. Customers who have an LLM operating on Cloudflare’s Workers are eligible to utilize the Firewall and receive complimentary protection from this new feature, which aims to counteract emerging threats like prompt injection and other attack vectors.

To obtain sensitive information from an LLM, prompt injection attacks meticulously create a question that manipulates the model’s functionality and causes the AI to say exactly what the attacker needs. By providing the LLM with new orders, an attacker can use prompt injection to override prior instructions and perhaps force the LLM to divulge sensitive data or disable vital functions.

Cloudflare’s New Protection Uses AI to Combat AI

Cloudflare’s Defensive AI provides a granular method for protecting enterprise networks from emergent threats, such as AI-enhanced attacks, by employing artificial intelligence to identify suspicious emails, code, and traffic patterns.

As stated by the organization, it has broadened its range of services to include the training of AI models using customer-specific traffic patterns in order to customize a defense strategy according to the organization’s network and environment’s baseline behavior.

“By understanding ‘normal baselines’ in a customer’s environment and mitigating the threats that will move the needle towards increased resilience, Defensive AI is the crucial edge defenders need to stay ahead of today’s adversaries,” said Prince.

Given the advancements in AI like OpenAI’s ChatGPT, cybercriminals have become increasingly sophisticated in their phishing scams, which involve deceiving users into disclosing sensitive information through emails or messages. In the past, scams were not always very convincing due to written messages filled with errors like grammar mistakes or poor design. With the help of AI, hackers can now customize their emails to recipients, increasing the chances of convincing them to disclose passwords or sensitive data.

Cloudflare’s Defensive AI helps detect email threats faster, allowing scam messages to be identified before employees are tricked, according to the company.

Cloudflare is creating an application programming interface anomaly detection model as part of the same defensive strategy to stop network intrusions, application attacks, and data theft. The goal is to create models of normal network activity and then monitor traffic anomalies. As attacks stray from the appropriate application behavior, these models will defend against malevolent attacks.

“We have entered into an era where fighting AI with AI is non-negotiable. A personalized approach to protect data and defend against complex threats unique to an organization’s attack surface – at speed, and scale – is paramount,” added Prince.