The Rise of AI-driven Cyber Attacks


It is hard to deny the adaptability that AI has provided cybersecurity intruders as a number of high profile crimes has occurred with multipronged attacks across the tech stack as well as towards human employees. The biggest AI-driven cyberattacks in 2025 include AI-powered phishing and social engineering, sophisticated malware that can adapt in real-time, and the use of deepfakes for impersonation. Other major threats involve AI being used for large-scale disinformation campaigns and the creation of malware that can automatically evade detection. 

Major attacks and incidents reported in 2025 include:

The Arup Deepfake Fraud: In a truly surprising incident reported in August, cybercriminals used AI-generated deepfakes of senior executives in a video call to convince a finance employee at British engineering firm Arup to transfer $25 million. The British multinational design and engineering company behind world-famous buildings such as the Sydney Opera House fell prey to this malicious AI fraud. The employee contacted was fooled by the realistic video and voice clones of actual company leaders, underscoring AI's use in advanced social engineering.

Claude Chatbot Exploitation: A report from Anthropic in August detailed how a hacker manipulated the AI chatbot Claude to automate every stage of a campaign against 17 different organizations. The attacker prompted Claude to perform reconnaissance, generate custom malware, and write personalized ransomware notes, essentially allowed it to fully automate and scale cybercriminal operations Targets included a bank, healthcare providers, and a defense contractor demanding ransoms of up to $500,000.

Salesforce Agentforce Compromise: In 2025, attackers used an indirect prompt injection to compromise Salesforce's Agentforce AI. The attacker would embed malicious instructions within a legitimate customer interaction, such as a "Web-to-Lead" form. This form data, including the hidden payload, would then be stored in the Salesforce database. By submitting ordinary data, they were able to force the autonomous agent to perform unauthorized commands, creating a risk of mass customer data exfiltration. Luckily, Salesforce responded by patching the system and adding security controls before any large data breaches occurred. 

It is critical for the security function of any firm to stay vigilant but also to keep abreast of new innovations across the technology landscape and understand that a dexterous approach is needed in these evolving times.

Comments

Popular posts from this blog

To SWOT or Not?! Answer = Technical SWOT

M&A Deals: A Bullseye for Cyberattacks