Article

My answer: Yes

AI in cybersecurity is somewhat overhyped, (enven if AI remains a valuable tool in certain areas). The idea that AI alone can secure networks, predict every threat, and replace human analysts is often exaggerated (and just cleary not possible). While AI provides significant benefits in certain aspects of cybersecurity, its limitations and the need for human oversight make the hype around it unrealistic in many cases. But, let's take a deep look at it.

List of reasons Why AI in Cybersecurity is Overhyped:

1. Over-reliance on Automation and AI-Driven Solutions

AI is often marketed as a silver bullet for cybersecurity, capable of fully automating security processes (that can be true but, in many case just a random bullshit). However, AI-driven solutions are not infallible and may struggle with emerging or sophisticated threats. For instance:

• Static Models: AI models are trained on past data and may not be able to detect entirely new attack vectors, such as zero-day vulnerabilities, where no prior data is available.
• False Positives and Negatives: AI systems can produce false positives (flagging benign activity as malicious) and false negatives (missing actual threats), leading to inefficiencies and security risks. It still requires human intervention to verify these findings and make context-driven decisions. In that case the real question is: why do we need AI if we still need human intervention?
• Limited Contextual Understanding: AI can analyze large volumes of data quickly, but it lacks the deeper contextual understanding of threats that human analysts bring. Cybersecurity threats often have a social or behavioral component (e.g., phishing scams) that AI struggles to comprehend fully without context. And in that case, that mean that to work, each AI model need to be trained with a lot of custom data of where their are installed, and that's not always possible or cost too much.

2. AI Itself is Vulnerable to Attacks

Ironically, AI models used in cybersecurity can be manipulated by attackers through adversarial attacks. Hackers can subtly alter input data (like network traffic or malware samples or just by exploiting weak model) to deceive AI systems into misclassifying threats. This demonstrates that AI can’t be entirely trusted to protect itself, let alone a complex security infrastructure. Moreover, AI models can be reverse-engineered by attackers, giving them insight into how to evade detection by the system.

3. Data Dependency and Quality Issues

AI requires vast amounts of high-quality data to function effectively. However, the data used to train AI models can often be incomplete, outdated, or unbalanced:

• Data Scarcity: Especially for novel attacks or niche vulnerabilities, there may not be enough data to train an AI model effectively (have fun to make an AI on industrial control system lol).
• Biased Data: AI can inherit biases from the training data. For example, if an AI system is trained predominantly on malware from a specific region or type of threat, it might overlook other types of attacks from different regions, leaving gaps in protection.
• Data Privacy: Leveraging large datasets for AI also raises concerns about privacy and regulatory compliance, especially with laws like GDPR (CNIL ❤️). The management and processing of sensitive data may create vulnerabilities and the source of the data can be a problem too.

4. Cost and Expertise Requirements

Despite claims that AI can reduce costs, implementing AI-driven cybersecurity systems is expensive and complex.

• High Costs: Building, deploying, and maintaining AI systems require significant investment in terms of infrastructure, expertise, and data management. Small and medium-sized businesses, for example, might find these costs prohibitive.
• Skill Shortages: The expertise required to manage AI in cybersecurity is specialized. There is already a well-documented shortage of skilled cybersecurity professionals, and adding AI to the mix requires even more niche knowledge of data science and machine learning, further complicating workforce issues.

5. Hype Overshadows the Human Element

The marketing narrative around AI in cybersecurity often suggests that AI will replace human cybersecurity professionals (fun fact hackers vs AI, AI always loose xD), which is misleading. While AI can automate repetitive tasks (like scanning logs or filtering alerts), it cannot replace the critical thinking, intuition, and decision-making capabilities that human analysts possess. Many cyberattacks involve human behavior, such as social engineering, which AI struggles to interpret and respond to effectively.

• AI can enhance human decision-making, but it cannot fully replace human judgment in dynamic and nuanced situations.

6. Not a "One Size Fits All" Solution

AI tools are not universally applicable to all areas of cybersecurity. Different organizations face different types of threats based on their size, industry, and digital footprint. AI may be overhyped as a general solution, but in practice, it needs to be customized and fine-tuned to specific environments to be effective. This complexity is often downplayed in favor of selling AI as a ready-made solution for every cybersecurity issue.

7. Inability to Handle Sophisticated Threats

Cyberattacks are evolving at a rapid pace. While AI can handle well-known and patterned attacks (like phishing or malware detection), it struggles with more sophisticated, multi-stage attacks involving advanced persistent threats (APTs), which can adapt, change their behavior, and evade detection over long periods. These complex attacks require human intelligence and strategic intervention, areas where AI is often inadequate.

8. Ethical and Legal Concerns

The increasing use of AI in cybersecurity raises ethical and legal concerns:

• Accountability: If an AI system fails to detect a breach or inadvertently causes a data leak, who is responsible? The legal and ethical frameworks surrounding AI in cybersecurity are still being developed, adding complexity to its widespread adoption.
• Privacy Risks: AI systems need to process vast amounts of data to be effective, but in doing so, they might inadvertently infringe on user privacy or violate data protection laws.

---

Conclusion

While AI has undeniable potential in cybersecurity, particularly in automating repetitive tasks, detecting patterns in large datasets, and providing predictive insights, the current level of hype surrounding AI often overstates its capabilities. AI is not a panacea for cybersecurity challenges. Its effectiveness depends on the quality of data, human oversight, and the ability to adapt to an evolving threat landscape.

In practice, AI should be seen as a complementary tool—augmenting, rather than replacing, human expertise. The notion that AI alone can safeguard complex systems against sophisticated cyber threats is unrealistic and inflated, hence the argument that AI in cybersecurity is, indeed, overhyped.