Follow Us on Google Discover
AI-driven cyberattacks increased by 56% in the past year, with more than 25% of organisations that experienced a malicious attack reporting that the incident was driven by artificial intelligence.
This is according to IBM’s 2026 Cost of a Data Breach Report, which found that AI-driven attacks now cost businesses an average of $6.04 million per breach.
IBM said attackers are increasingly using AI to automate attacks, identify vulnerabilities and operate at greater scale, reducing the time needed to exploit weaknesses and increasing the potential financial impact on businesses.
What IBM is saying
IBM found that more than one in four organisations experienced a malicious AI-driven attack, representing a 56% increase from the previous year.
Also Read
- “More than one in 4 organizations experienced a malicious, AI-driven attack—a 56% increase over last year,” IBM said.
- The report also found that AI-driven attacks cost about $1 million more per breach, with such attacks costing an average of $6.04 million, compared with $5.03 million for malicious attacks that did not involve AI.
Deepfake impersonation accounted for 45% of AI-driven attacks, making it the most common form identified in the study.
AI-generated malware accounted for 19%, while AI-generated phishing and other communications accounted for 17%.
IBM said generative AI has made social engineering attacks cheaper to create and harder to detect, allowing attackers to produce convincing content and manipulate victims at greater scale.
- “Attackers are abandoning human speed for machine speed,” IBM said, adding that generative AI has reduced the time, cost and expertise required to launch attacks.
Hackers are now targeting AI systems
Criminals are also increasingly targeting the AI systems that businesses are deploying.
- IBM found that 21% of organisations experienced a security incident involving an AI model or application, up from 13% a year earlier.
- The most expensive incidents involved model inversion, which cost an average of $6.07 million, and prompt injection, which averaged $5.89 million.
- Model inversion involves attempts to extract sensitive information from an AI model, while prompt injection involves manipulating an AI system through malicious instructions.
IBM said these vulnerabilities are particularly significant as AI becomes increasingly connected to corporate applications, data and workflows.
More insights
The growing use of AI in cybercrime is also becoming a major concern across Africa, with artificial intelligence enabling 55% of reported cybercrimes on the continent, according to INTERPOL’s African Cyberthreat Assessment Report 2026.
- The 40-page report also found that financial losses linked to cybercrime more than doubled, rising from $192 million in 2024 to $484 million in 2025. The findings are based on survey data from 36 African member countries.
- INTERPOL said the trend reflects a broader shift in the continent’s cybercrime landscape, with criminal operations moving from isolated attacks to industrialised and borderless networks increasingly powered by AI tools.
Neal Jetton, Director of INTERPOL’s Cybercrime Unit, said the scale and sophistication of the threat require countries to move beyond isolated national responses and strengthen coordinated cross-border action.
What you should know
- The growing financial impact of AI-enabled cybercrime is also reflected in data from the United States, where the Federal Bureau of Investigation (FBI) recorded more than $893 million in losses linked to AI-related cybercrime in 2025.
- The figure was disclosed in the FBI’s 2025 Internet Crime Report, published by the Internet Crime Complaint Center (IC3), which tracks trends in internet-enabled crimes and the financial losses reported by victims.
The report said criminals are increasingly using AI to create convincing synthetic content, including fake social media profiles, voice recordings and videos, which can be used to impersonate individuals, deceive victims and facilitate fraud.
