By Sohil Ramdas | Trace3 Innovation Principal
In a relatively short span of time, artificial intelligence has transformed how businesses approach technology. But how exactly has this happened?
Developers are in a new era of writing code with the help of AI coding assistants. Employees are generating content, building presentations, and performing deep research with their favorite Chatbots. Agents are autonomously planning and executing on behalf of company resources to allow more flexibility in how we operate. The scope of AI aims to be limitless, and that’s an exciting thought for both today and tomorrow.
But while there is no shortage of use-cases for AI enablement, the flip side of the coin has been this technology advancement has also unveiled a brand-new attack surface. Cybersecurity departments are now responsible for governing the utilization of shadow AI, providing guardrails for applications using large language models and ensuring Agents operate in the capacity they were meant to be. Companies are in a competitive business landscape to adopt AI, but every new trend provides a potential new risk.
So, what has been the response? The answer to these challenges has proven to be a balance between best of breed pioneers and comprehensive platform plays. Several security start-ups have come to the table with focused efforts on the specific attack vectors and concerns. Additionally, established vendors have aimed to both enhance their own products and consolidate the space by acquiring technologies. Together, this has created the AI security market that provides cybersecurity teams with multiple approaches to the emerging space, offering a choice on when and how to invest.
To further complicate matters, the AI adoption and associated security controls must be an on-going assessment for an organization. The dynamics are changing daily, and the capabilities are growing exponentially. While this is significant for innovation, it can be a tough task to build a secure strategy for implementation. The question then becomes: How does a company embrace and scale its AI footprint while ensuring its Cybersecurity teams have the necessary solutions in their toolbox to address the risks?
AI Security Snapshot
Chances are the employees of most organizations have used Generative AI in some capacity. Some employees may even use it as frequently as they check their email or respond to a ping. However, what may not be as well-known are the security concerns and blast radius that Generative AI introduces. Here is a quick recap of a few of those key risk areas:
-
Shadow AI: Third-party chatbots and AI-based applications increase employee efficiency but without proper visibility and control, the potential for data leakage increases.
-
LLM Guardrails: Generative AI applications that utilize LLMs are susceptible to attacks such as Prompt Injection, Data Poisoning, Output and Data Leakage.
-
Agentic AI: The ability for agents to utilize tools and execute based on their decisions provides opportunities for privilege escalation, data breaches, and hallucinations.
-
MCP Security: Requests from MCP clients to servers introduce concerns around permissions, secret exposure, and prompt injection.
This is by no means an exhaustive list, and every new advancement introduces both a potential new risk and a company looking to solve that challenge.
For a deeper dive into the world of AI Security and MCP Security, read the following Trace3 blogs: “4 Words to Supercharge Your GenAI Security Strategy” and “Unpacking MCP’s Security Challenges and the Defenses Rising to Meet Them.”
How Security Companies Reacted
We have established AI is a company-wide use-case that expands the blast radius with different vectors of attack, data concerns, and governance issues. What are the solutions?
From one angle, the start-up community has rapidly identified the problems and developed products to address the risks. Often these products are built with a niche purpose in AI security (e.g., LLM Guardrails, MCP Gateways, or AI Redteaming). The pioneers in this space aim to operate in parallel with the still-growing AI market. With their domain-specific expertise, agility in development, and flexibility to changing conditions, these solutions allow enterprises the ability to deploy mitigating controls and technologies as required.
From another angle, there have been strategic efforts to consolidate the market by major technology companies. This shift is occurring during the early stages in the start-up world as vendors recognize the market and aim to strengthen their own security portfolio. This allows companies to gain synergies between their existing capabilities (e.g., EDR) and AI specialties (e.g. Shadow AI). In this way, they can offer a proven AI security platform as a feature in their product suite sooner while continuing to build out their own capabilities. Just in 2025 alone there have been a few key acquisitions such as:
-
Check Point acquisition of Lakera for LLM protection
-
CrowdStrike acquisition of Pangea for AI application governance
-
Palo Alto Networks acquisition of Protect AI for overall AI security
-
F5 acquisition of CalypsoAI for AI model security
-
SentinelOne acquisition of Prompt Security LLM protection and AI governance
Both paths, best in breed solutions for today’s AI security and the consolidated platform plays for an expanded footprint, have provided organizations a way to address the risks specific to their environment.
Balancing Innovation and Security
Finally, now that we have enough information around the AI use-cases, security concerns, and the potential solutions for those risks, we can address the foundational question: How does a company embrace and scale its AI footprint securely?
The answer resides within a self-assessed decision framework for each enterprise. While every company will operate in its own timeline, a similar evaluation will occur across the following points:
AI Timeline
-
Early Adoption/Innovation: This is when experimentation across the organization is abundant. Shadow AI and developer-led LLM use-cases lead the way. Pioneers in the market offer the path to quick visibility and guardrails.
-
Scaling: AI is now across the organization and being used/developed at different speeds. A hybrid approach provides a perfect opportunity to address critical risks while developing governance, policies, and procedures.
-
Production: The company has operationalized the AI ecosystem with model usage, agent development, and pipelines deployed throughout. Platforms provide synergies across toolsets and consistency for enterprise-wide policy enforcement.
Security Focus
-
Priority Issues: Pioneers are quick to deploy and deliver specialties when data leakage, prompt injection, and similar risks are prevalent and immediate.
-
Governance and Visibility: Platforms can integrate from the top-down to provide centralized oversight and controls for Cyber teams.
Considerations
-
The amount of pioneer solutions to hit the market can be tough to evaluate and attempts around proof of concepts are resource intensive.
-
Integrations for platforms may not align with the speed an organization adopts AI. Additionally, if the technology isn’t already in-house, costs could be prohibitive.
Technology Outlook
-
Pioneers: Their speed to market and product-focused solutions (e.g., LLM guardrails) are major benefits to a Cybersecurity team.
-
Platforms: This can reduce friction in terms of implementation into their environment and allow for a broader and quicker deployment of security controls.
-
Hybrid: This provides the ability for an organization to securely innovate while establishing the framework for its AI evolution.
In the end, each path, embracing new technologies or expanding the current tech stack, will have its own merits and potential limitations. A model focused on a combination of both pioneers and platforms will allow organizations to remediate immediate risks while maturing their AI ecosystem. Enterprises will find their most successful strategy centers around a constant assessment of their landscape, security concerns, and timelines to determine when and how to invest. This, in turn, will provide Cybersecurity teams with the required intel to prioritize the risks and ensure their toolbox is properly updated.
If you’re curious to learn more or want to stay on top of the latest developments in Innovation, feel free to reach out to us at innovation@trace3.com.
Sohil Ramdas serves as an Innovation Principal on Trace3’s Innovation Team. With a career background in Cybersecurity, he leverages his extensive experience in both industry and consulting roles to now provide clients with guidance on emerging technologies. His objective is to supply organizations with the expertise required to securely innovate and scale in a rapidly evolving technology landscape.