In the rapidly evolving landscape of cybersecurity, AI applications have emerged as the new frontier for threat actors. At RSA Conference 2025, Akamai Technologies unveiled its latest security innovation, Firewall for AI, designed specifically to combat the unique threats facing generative AI (GenAI) and large language models (LLMs). We sat down with Rupesh Chokshi, Senior Vice President and General Manager of Application Security at Akamai, to talk about these emerging threats and Akamai’s strategic response.

The Changing Threat Landscape

Highlighting current challenges, Chokshi shared findings from Akamai’s latest State of the Internet (SOTI) report, which recorded a 33% surge in web application attacks, reaching 311 billion.  Nearly half of these attacks—approximately 150 billion—specifically target APIs. Layer 7 DDoS attacks have also seen significant growth, with commerce, finance, and high-tech sectors being the primary targets.

These trends emphasize the urgent need for advanced defenses designed for the evolving AI-driven threat environment.

As Chokshi explained, “There’s increasing adoption of AI across every vertical, industry, and government agency, each trying to figure out how to do good—or do better—by leveraging AI.” This widespread adoption has created new security challenges that traditional web application firewalls aren’t designed to address.

Unique Threat Vectors for AI Applications

Chokshi outlined how traditional web application firewalls fall short when confronting new threats like prompt injection, data poisoning, and multilingual exploits that target LLMs. These attack vectors can bypass existing safeguards, exploit sensitive data, and introduce misinformation through AI models. In response to these growing threats, Akamai developed the Firewall for AI to secure AI traffic, including prompts and responses, against these evolving risks.

The security challenges facing AI applications differ substantially from conventional web application threats. Chokshi highlighted several key attack vectors:

Prompt Injection: Attackers are reframing prompts and employing role-play techniques to bypass guardrails, extract sensitive information, or inject malicious content such as phishing websites.

Data Poisoning: Malicious data is being injected into training sources, potentially leading to compromised responses. For example, phishing URLs inserted into training data could result in chatbots directing users to fraudulent websites.

Sensitive Data Leakage: AI systems are susceptible to exposing PII and other confidential information.

Multilingual Exploit Chains: Attackers are using combinations of prompts in different languages to circumvent security measures.

These threats align with OWASP’s Top 10 for LLM security, which has begun to establish industry standards for securing AI applications.

Introducing Firewall for AI

Akamai’s new Firewall for AI addresses these emerging threats by securing AI traffic end-to-end, covering both inbound and outbound prompts. The firewall can be deployed at the edge or via API, with minimal disruption to existing systems.

Chokshi highlighted real-world use cases, such as a mortgage and real estate firm that used the firewall to detect and block 5 to 6% of risky AI interactions, including toxic prompts and data leakage. “The threats are real, and we believe we can help customers in many different ways as they continue their journey of adopting more GenAI capabilities to accomplish what they need,” says Chokshi.

“We have visibility at the Akamai edge, and we want to place the bump in the wire, focusing on providing capabilities for all AI traffic—whether inbound or outbound,” Chokshi noted.

The firewall offers detection for prompt injection, data leakage, and reputational risks, with responses ranging from alerts to traffic modification or denial based on the severity of the threat.

One of the key advantages of the solution is its deployment flexibility. Existing Akamai customers can integrate the firewall into their current traffic flows, while new customers can implement it through API integration.

API LLM Discovery: Enhanced Visibility

Complementing the firewall, Akamai has also introduced API LLM Discovery, which provides visibility into AI and LLM usage within organizations. This capability helps security teams identify unauthorized or previously unknown AI interactions within their environments.

“Many customers didn’t realize that GenAI and LLMs were already being utilized within their organization until we showed them dozens of API calls taking place,” Chokshi revealed.

This visibility is critical for organizations in regulated industries that must comply with emerging AI governance frameworks such as Gartner’s AI TRiSM, OWASP’s Top 10 for LLMs, NIST guidelines, and the EU AI Act.

Addressing the Hybrid Reality

Chokshi emphasized that most enterprises will operate in a hybrid environment for the foreseeable future, maintaining traditional applications while gradually introducing AI-powered capabilities. Akamai’s approach acknowledges this reality by providing protection that works across both conventional and AI workloads.

Even for “internal” AI applications, such as those used by customer service departments or employee-facing systems, the security risks remain significant. As Chokshi pointed out, “If you start sharing misinformation within the company, you may ultimately end up passing some of it on to external customers.”

Looking Ahead

Akamai’s strategy involves two parallel tracks: using AI to enhance security capabilities (as they’ve done for years with their adaptive security engine, bot mitigation, and API behavioral engine) and developing new products specifically focused on securing GenAI, LLMs, and AI applications.

The Firewall for AI represents Akamai’s initial entry into this space, with more comprehensive AI security platforms planned for the future.

As AI adoption accelerates across industries, the security challenges will only grow more complex. Traditional security approaches are insufficient for protecting these new systems from sophisticated attacks. With its Firewall for AI and API LLM Discovery capabilities, Akamai is positioning itself at the forefront of AI security, providing the visibility and protection that organizations need to deploy AI safely.

For security professionals, understanding these new threat vectors and implementing appropriate safeguards will be essential as we navigate this new frontier of cybersecurity challenges.

Guest: Rupesh Chokshi
Company: Akamai
Show: Secure By Design

This summary was written by Emily Nicholls.