AI-powered cyber-attacks in 2025 and how to fight back

Is AI making cybercrime worse? The answer to the question is a resounding yes! 

AI-powered cyberattacks are proliferating, targeting both businesses and individuals alike. This makes it critical for enterprises to adapt and fortify their infrastructure to defend against generative AI (GenAI) threats.

GenAI in cybersecurity uses AI models, including those based on large language models (LLMs) or generative adversarial networks (GANs), to create or manipulate data for defensive and offensive purposes.

Defensive use cases:

• Automated responses: GenAI can create automated responses to incidents. This includes generating patches or containment strategies to accelerate incident response processes.
• Data augmentation: Synthetic data generated by GenAI systems can train security models. This can be a real asset for security teams when real-world data is limited. This approach helps improve the accuracy of threat detection systems.
• Threat simulation: GenAI can help security teams create realistic attack scenarios to test and train cybersecurity systems. This approach allows organizations to identify potential weaknesses before threat actors exploit them.

Offensive use bases (by attackers):

• Data forgery: Hackers can use GenAI to create fake data or credentials to bypass authentication systems. They also use this technique to poison datasets used by defensive AI tools.
• AI malware: Threat actors can generate new, unique malware variants by altering code structures to evade traditional signature-based detection systems.
• Social engineering: GenAI can make social engineering attacks harder to spot. Examples of AI-powered hacking include compelling phishing emails, texts, or deepfake content (including fake videos or voice recordings).

A recent example of a GenAI social engineering attack is the impersonation of the likeness of YouTube CEO Neal Mohan. In this security event, cybercriminals targeted content creators with AI-generated deepfake videos featuring Mohan. These deepfake private videos aimed to install malware and steal credentials.

Other AI cybersecurity risks for businesses include password cracking, automated attacks (including Distributed Denial of Service or DDoS), and vulnerability scanning that leads to exploits.

But not everyone is ready.

According to a study conducted by the World Economic Forum, as much as 66% of organizations surveyed expected AI to have a massive impact on cybersecurity this year. Yet only 37% stated they had established processes to evaluate AI tools before implementation.

 The good news is that this is slowly changing. According to the Worldwide Security Spending Guide from the International Data Corporation (IDC), global security spending will rise by 12.2% year on year in 2025 to meet the demands of increasingly frequent and complex cyber threats enhanced by GenAI.

Western Europe and the United States represent the biggest chunk, with over 70% of global security spending. Still, every part of this planet is expected to increase consistent security spending this year.

Security teams must use AI-based cybersecurity tools to detect anomalies quickly and respond to threats in real time. This includes monitoring for unusual network traffic or behavior that might indicate the presence of adaptive malware or automated exploits. 

Use AI detection tools to analyze suspicious videos, voices, or documents. Use honeypots or decoy systems to mislead AI-powered attacks, slowing them down and giving your team time to respond.

Encryption defends against AI-powered attacks by making data inaccessible or unusable to attackers. Even if hackers use advanced AI tools, encryption:

• It prevents data exploitation by making data unreadable without a decryption key.
• Mitigates the risk of deepfakes as encrypted communication channels prevent AI-generated deepfakes or intercepted communications from being tampered with or convincingly faked. Attackers can’t access the plaintext data to mimic.
• Protects against data poisoning as encrypting stored data makes it difficult for hackers to alter or inject malicious data into training data sets.
• Hinders automated exfiltration even though AI can automate data theft at scale. For example, even if an AI-powered botnet breaches a system, encrypted files require significant computational resources to crack, giving incident responders time to respond.
• Secures data in transit, thwarting AI tools that analyze patterns in communications.
• Limits the impact of credential theft as the encryption of stored credentials (such as hashing with salts) means stolen password databases are more complex to exploit. Even if credentials are cracked, encrypted systems often require additional authentication.

Thoroughly vet vendors and partner with those that follow ethical AI practices and data security protocols. Always review the data retention policies of any AI platform your team uses and ensure tools don’t leak or store your sensitive business data externally.

Establishing clear guidelines on how employees can use AI tools such as ChatGPT, Copilot, or generative platforms is important. Restrict the use of generative AI for sensitive or proprietary data, or don’t use them at all. Educate and ensure employees understand data privacy laws when using AI platforms.

Conduct regular security training workshops and teach staff to identify AI-generated phishing emails and manipulated content. Always instill security best practices, including password hygiene and MFA.

As AI-powered tools rapidly evolve, so do the threats that come with it. From deepfakes to data breaches, organizations must take a proactive approach to defend against generative AI threats.

By combining smart policies, robust cybersecurity, employee awareness, and ethical AI usage, you can fortify your defenses against AI-powered cyberattacks. Staying informed and adaptable is not critical to business continuity. It’s the only way to survive and grow in the modern digital landscape.