I created Custom GPTs to serve multiple needs: generate better content, support personal branding efforts, or improve investment decisions. I recently started to work on a new project, customizing a GPT to act as an AI security assistant for small business owners.
Custom GPTs trump prompt engineering. In the past year, there was no day without a post on prompt engineering in my LinkedIn feed. And it’s understandable because prompt engineering helps us get the most out of Generative AI. But Custom GPTs take things to the next level. A custom GPT is more effective because it better integrates specific contexts and needs directly into the model. It provides tailored and relevant responses without the need for complex and ongoing prompt adjustments. It’s much more scalable and can be used by individuals less skilled in prompt engineering.
Small businesses are in a unique cybersecurity position. They are exposed to an increasing array of cyber threats but have limited competencies and usually no dedicated security team. The consequences of cyberattacks can be significant, ranging from financial damage to the loss of customer trust, and ultimately, going out of business. This is where gen AI can help, offering solutions previously inaccessible to smaller companies, among others in the form of artificial intelligence security assistants.
These specialized chatbots represent a new breed of defense: powerful always-on partners working with business owners to protect their businesses. They can learn, adapt, and evolve – providing insights tailored to the particular context of each business. Exactly as an expert security consultant would do but 24/7 and at a much lower cost. It makes expensive security services available to any business, regardless of size or tech expertise.
An AI Security Assistant for Small Businesses
When designing an AI Security Assistant for Small Businesses, there are several important considerations:
User-Friendly Interaction
Given that the AI will primarily provide consultancy and support, it’s crucial to customize a GPT that is intuitive and user-friendly for small business owners who may not have advanced technical expertise. The security copilot should use clear, understandable language and provide step-by-step guidance for making decisions, such as setting up security measures or identifying potential security risks.
Customizable Advice
The AI assistant should offer advice tailored to the unique needs and risk profiles of different small businesses. This involves providing sector-specific advice, from retail to services, understanding the common threats in each area, and offering strategies to mitigate these risks. The assistant should be able to guide businesses in developing a cybersecurity plan that includes elements like employee training, data protection measures, and incident response strategies.
Up-to-date Security Information
The Machine Learning model should be equipped and regularly updated with threat intelligence capabilities, keeping small businesses informed about the latest cyber threats and scams. This includes not just general information but also alerts and advice on immediate actions to take in response to emerging threats. The assistant should be able to explain complex threats in simple terms, helping business owners understand the implications and the necessary steps to protect their operations.
Educational Content and Resources
Beyond immediate consultancy and support, the AI assistant should provide educational content and resources to small businesses. This includes information on best practices for cybersecurity, how to recognize and avoid common scams, and updates on regulatory compliance requirements. The goal is to empower business owners and employees with knowledge, turning them into active participants in their cybersecurity defense.
CyberPal – My AI Security Copilot experiment
Customizing a GPT Model to act as a Security Advisor for Small Businesses is a detailed process that focuses on the key elements mentioned above: behavior, security best practices, and providing up-to-date security information.
Here’s the process I followed to create my own AI security assistant, CyberPal:
Customizing the Model’s Behavior
Step 1: Define the Scope
Outline the security advisory roles the GPT model will fulfill: guide secure practices, advise on data protection, and identify security incidents (like scam attempts or ongoing security incidents)
Example Prompt: “Configure the model to identify the primary security concerns for a wide range of small businesses (from retail to IT services). The model should ask about the business type, size, and main digital assets to tailor its advice. The model will provide insights into best practices for securing a small business and will offer support to identify potential security incidents.”
Step 2: Tailor the Communication Style
Customize the model to communicate complex security concepts in terms suitable for small business owners without technical backgrounds. Use examples that reflect this simplified, clear communication style.
Design a conversational flow that mimics a human security consultant. ChatGPT has the tendency to shoot solutions after the first prompt. Instruct the model to ask clarifying questions before offering advice, ensuring the guidance provided is contextually relevant to the specific query.
Example Prompt: “Adjust the model’s responses to explain cybersecurity concepts (like encryption and two-factor authentication) in simple terms, suitable for someone with no technical background. Include analogies and examples relevant to everyday business operations. Program the model to follow an inquiry-based approach, where it first gathers information about the user’s specific situation through a series of questions before providing any recommendations.”
Incorporating Security Best Practices
Step 1: Fine-tune the model with Security Best Practices
Gather a comprehensive collection of cybersecurity best practices specifically tailored for small businesses. There are plenty of guidelines designed by reputable sources. I included guidelines from the National Institute of Standards and Technology (NIST) or the Cybersecurity & Infrastructure Security Agency (CISA).
Example prompt: “Integrate the attached set of security best practices into the model’s training. Ensure the model can relay these practices in a contextually relevant manner to different business scenarios.”
Step 2: Scenario-Based Training
Incorporate various security scenarios and case studies into the model’s training regimen. You can go two routes here: generic scenarios or industry-specific. The more scenarios, the better the model. This can help the AI provide more practical, actionable advice that small businesses can implement.
Example Prompt (does not include scenario details): “Train the model with a variety of cybersecurity scenarios, such as a ransomware attack on a small accounting firm or a phishing attempt targeting a local restaurant. The model should guide the user through the steps to mitigate the threat, emphasizing preventative measures and immediate actions.”
Providing Up-to-Date Security Information
Step 1: Curate a Knowledge Base
Develop a knowledge base that includes the latest security threats, vulnerabilities, and trends. While direct integration with live threat feeds may require significant investment, regularly updating this database manually can keep the information current.
Use reputable sources of cybersecurity updates: Government and Industry reports, specialized cybersecurity news websites, blogs, and forums.
Step 2: Regular Model Updates
Schedule periodic updates to the GPT model to incorporate the latest security information from your curated knowledge base. This process will ensure that the model will be able to provide up-to-date, specific advice in regard to the latest evolution of the threat landscape. It’s an ongoing effort that requires manual work and I’m looking for ways to automate this process.
Closing thoughts
Developing a Custom GPT AI Security Assistant tailored for small businesses like CyberPal, requires a thoughtful approach that balances ease of use, practicality, and up-to-date knowledge. By focusing on customizing the model’s behavior to suit non-technical users, incorporating fundamental security best practices, and maintaining a curated knowledge base of the latest threats and trends, small businesses can have access to a powerful tool that enhances their cybersecurity posture.
Custom GPTs help democratize access to cybersecurity expertise. It requires a single security professional to configure and update the model. Then this can be used by a large number of small business representatives with little knowledge of cybersecurity or prompt engineering. It’s a very effective way to empower businesses to uplevel their defense against common and sophisticated threats.