Artificial intelligence has moved beyond being a promise and has become a cross-cutting technology transforming how we work, innovate, and interact with information. However, the same potential that is revolutionizing business productivity is also altering the cybersecurity threat landscape. In recent months, experts have warned that AI is not only making cybercriminals' work easier, but is also multiplying their ability to launch faster, cheaper, and harder-to-detect attacks.
The latest data confirms this trend. According to the latest Check Point Research report, the average number of weekly cyberattacks per organization worldwide reached 2.336 in July, a 16% increase compared to the same month last year. In Spain, companies recorded an average of 2.068 weekly attacks, 1% more than in 2025. Although the growth is more moderate than the global average, the pressure on Spanish organizations remains intense, especially in sectors such as education, government, and consumer goods and services.
Record numbers of AI-powered cyberattacks
A report by Check Point Research, which analyzes cyberattack activity worldwide, reveals that education remains the most targeted sector , with an average of 4.848 attacks per week per organization, a 14% increase over last year. This is followed by government institutions, with 3.044 attacks (+11%), telecommunications (2.927, +6%), energy and utilities (2.759, +20%), and hospitality, travel, and leisure (2.614, +28%). In percentage terms, the business services sector accounted for 32,5% of reported victims, followed by industrial manufacturing (14,4%) and consumer goods and services (13,4%).
These data reflect that cyber risk is accumulating on multiple fronts simultaneously , as researchers point out. Not only is the volume of attacks increasing, but the entry points are also diversifying. Email remains one of the main vectors: one in every 128 emails analyzed was classified as phishing, and an additional 20% fell into risk categories such as spam or graymail. Africa recorded the highest phishing rate (one in every 106 emails), followed by North America (one in every 117).
Ransomware, the fastest growing threat
Among all the threats analyzed, ransomware stands out for its spectacular growth. Reported ransomware attacks reached 964 cases in July , 49% more than in June and 87% more than in July 2025. This increase makes ransomware one of the most profitable methods for criminal groups, with an impact that goes beyond temporary disruption: it can cause direct financial losses, leaks of sensitive information, and reputational damage. Therefore, it is vital to implement recovery strategies for serious failures to mitigate these impacts.
By sector, business services were the most affected, accounting for 32,5% of incidents, followed by manufacturing (14,4%) and consumer goods and services (13,4%). Geographically, North America accounted for 45% of reported cases, while Europe ranked second with 28%. The most active criminal groups included The Gentlemen and Qilin, each responsible for 14% of the attacks reported during the period.
Generative AI, a new risk front
The report also focuses on the exposure resulting from the widespread use of generative artificial intelligence tools in businesses. One in 36 prompts or queries made to AI platforms was classified as high-risk due to its potential to expose sensitive information. Furthermore, 88% of organizations using these tools reported activity considered high-risk, and 22% of the messages contained potentially confidential information.
Personal data is the most exposed category, present in 70% of the organizations analyzed, followed by financial data and network and IT infrastructure, each at 68%. This phenomenon reflects a growing challenge: harnessing the potential of AI without compromising critical assets. In many cases, the risk arises when employees input documents, reports, or customer data into AI assistants without being fully aware of the security implications.
OpenAI warns: defenses must be strengthened
Meanwhile, OpenAI has published a report warning that AI is accelerating cyberattacks and proposing urgent measures for companies and organizations to strengthen their defenses. The company behind ChatGPT points out that the same models capable of assisting programmers and automating tasks can also be used to discover vulnerabilities, chain together security flaws, and accelerate attacks that until recently required specialized teams and significantly more time.
OpenAI recommends improving both traditional security controls and the speed at which vulnerabilities are detected and remediated. Among its suggestions are strengthening principles such as least privilege, defense in depth, network isolation, system updates, and continuous monitoring. It also advises using AI itself as a defensive tool, with agents capable of analyzing code, configurations, and documentation to identify risks before they are exploited. However, the company emphasizes that automation should be phased in and that high-impact decisions should remain under human supervision.
Spain prepares: the CCN calls for self-evaluation
At the national level, the National Cryptologic Center (CCN) has urged all entities included in the National Security Framework (ENS), especially public administrations, to complete a new self-assessment form by September 15 regarding their preparedness against threats arising from the offensive use of artificial intelligence. The agency warns that these threats are not hypothetical, but rather part of the current operating environment for public administrations and their technology providers.
The form, available on the INÉS platform, covers areas such as social engineering, internal use of generative AI tools, identity and credential protection, incident detection and response, document and communications integrity, vendor dependency, and staff training. The CCN emphasizes that the self-assessment should serve to identify the actual level of preparedness, not to declare a formal compliance position, and that the results should be integrated into the ENS improvement and audit plans.
A real case: the attack on Taiwan
Meanwhile, the Taiwanese government has reported an unprecedented security breach, attributed to AI-assisted cyberattacks. According to the Digital Affairs Administration (MDA), the attack compromised at least 85 government user accounts and extracted more than 2.500 personnel records before spreading to the Nuclear Security Agency and seven energy sector companies. The Israeli company Dream, which assisted in the detection, indicated that the perpetrators used open-source AI agents to create an autonomous hacking tool capable of operating as a coordinated team of cyber attackers.
This case illustrates the speed with which AI-supported attackers can achieve their objectives, although experts point out that there is still a person behind it all . Someone decides who to attack, sets the targets, and gives the directives. It's not completely autonomous, but the ability to automate drastically reduces the time and resources needed to launch sophisticated campaigns.
In short, artificial intelligence is redefining the cybersecurity landscape. Organizations that maintain slow and primarily manual security processes could be at a disadvantage against attackers capable of operating at machine speed. The advantage will lie with those that combine traditional best practices, automation, artificial intelligence, and human oversight. The race will not be between humans and machines, but between organizations capable of incorporating AI with a robust risk management strategy and those that adopt it without one. Cybersecurity is everyone's responsibility, and the time to act is now.
