A 2023 business survey revealed that 62% of enterprises have already implemented artificial intelligence (AI) for cybersecurity or are exploring additional ways to leverage the technology. However, as AI continues to advance, it also opens up new possibilities for sensitive information to be misused.

Globally, organizations are integrating AI and automated security measures into their systems to reduce vulnerabilities. As AI evolves, so do the threats it must contend with. A recent IBM report highlighted that the average cost of a data breach is now an eye-watering $4.45 million. The rise of generative AI (GAI) is expected to make automated AI-driven attacks more accessible, with a level of personalization that could make these threats harder for humans to detect without GAI’s help.

While AI refers to a broad category of intelligence-based technology, GAI is a specific subset that generates new content across various modes and even combines them. The main concern in cybersecurity is GAI’s ability to “mutate,” which means it can self-modify its code. When an attack driven by a model fails to breach a system, GAI can adjust its behavior to try again, making it even more difficult to stop.

The increase in cyberattacks comes alongside the growing availability of AI and GAI through tools like GPT, BARD, and various open-source platforms. Cybercrime tools such as WormGPT and PoissonGPT are suspected to have been developed using the open-source GPT-J model. While some GAI language models, like ChatGPT and BARD, have built-in anti-abuse features, the sophistication of GAI in crafting attacks, bypassing security defenses, and cleverly engineering prompts remains a significant concern.

This issue also ties into the larger challenge of distinguishing real information from fake. As the line between truth and falsehood becomes increasingly blurred, it’s crucial to ensure the accuracy and reliability of GAI models, especially when it comes to identifying fraudulent information. Using AI and GAI algorithms to defend against attacks generated by these technologies offers a promising solution.

Standards and Initiatives in AI-Driven Cybersecurity

A recent report from the Cloud Security Alliance (CSA) pointed out that generative AI models could play a vital role in scanning and filtering security vulnerabilities. The CSA highlighted how OpenAI and large language models (LLMs) could be effective at detecting potential threats by acting as vulnerability scanners. For example, an AI-powered scanner could identify insecure code patterns for developers to fix before they turn into a major risk.

Earlier this year, the National Institute of Standards and Technology (NIST) launched the Trustworthy and Responsible AI Center and introduced the AI Risk Management Framework (RMF). This framework helps AI developers and users understand and manage the risks associated with AI systems while offering best practices for mitigation. However, despite these efforts, the RMF still lacks some depth. In June, the Biden-Harris administration announced that a group of developers would create new guidelines to help organizations assess and address risks related to GAI.

As the cost of cyberattacks continues to drop with lower entry barriers, these frameworks will be crucial in guiding organizations. However, the increasing number of AI- and GAI-driven attacks will push developers and companies to quickly build on these foundations.

How GAI Can Benefit Cybersecurity

GAI can dramatically reduce detection and response times, making it an essential tool for addressing AI-generated attacks. Some key benefits include:

• Detection and Response: AI algorithms can analyze large datasets and track user behavior to detect unusual activity. GAI can go a step further by generating timely countermeasures or decoys to neutralize these threats before they escalate. This helps prevent intrusions that could sit unnoticed in an organization’s system for days or even months.
• Threat Simulation and Training: AI models can simulate realistic cyberattack scenarios, such as malware and phishing attempts, helping organizations prepare for potential threats. As GAI adapts and learns, these simulations become progressively more complex, improving internal systems and strengthening defenses over time.
• Predictive Capabilities: As IT networks evolve, AI can provide predictive analysis to assess shifting vulnerabilities. Consistent risk assessments and threat intelligence can support proactive security measures to stay ahead of emerging threats.
• Human-Machine and Machine-Machine Collaboration: While AI and GAI can significantly enhance cybersecurity, human intervention is still necessary. AI’s pattern recognition is advanced, but human creativity and decision-making remain vital. Collaborative efforts between humans and machines can help reduce false positives (incorrectly flagged attacks) and false negatives (missed threats). Additionally, machine-to-machine collaboration can further reduce errors, improving overall detection and response.
• Collaborative Defense: Human-machine and machine-machine collaborations can enable organizations to work together to enhance their defenses. By using cooperative game theory, organizations can model cyberattack scenarios, predict adversary actions, and determine the best defense strategies. This collaborative approach can strengthen cybersecurity policies and improve the effectiveness of defense efforts. AI systems designed to cooperate with other AI models across competing organizations could foster a more stable and cooperative cybersecurity environment.

A Modern Approach to Cybersecurity

The global market for AI-powered cybersecurity technologies is expected to grow at a compound annual growth rate of 23.6% by 2027. While it’s difficult to predict the exact future of generative AI in cybersecurity, it’s clear that AI shouldn’t be feared as a potential threat. A modern approach to cybersecurity involves standardizing AI models while fostering continuous innovation. This approach will allow businesses to stay ahead of emerging threats and take full advantage of AI’s potential to enhance security.