Axi · Сквозной — фон контента (skin) · 16 кат.Axi · Сквозной — фон контента (skin) · 16 кат.

The Number of Cyberattacks in 2026 More Than Doubled Amid AI Development

• Evgenia Sliv (edited: )
The Number of Cyberattacks in 2026 More Than Doubled Amid AI Development

The number of cyberattacks in 2026 has sharply increased: according to Bridgewater Associates, the number of incidents more than doubled in about three months. The acceleration began after the weights of the Chinese AI model GLM 5.3, which has advanced capabilities for working with code and identifying vulnerabilities, were released to the public at the end of August.

The fund has not yet proven that the release of this model was the cause of the surge in attacks. However, Bridgewater estimates the probability of an economically destabilizing cyberattack in the next year at about 1 in 3. Banks, payment companies, and cryptocurrency exchanges are at particular risk.

Why the number of cyberattacks has sharply increased

The chart from Bridgewater Associates was published on X by Charles Edwards, founder of Capriole Investments, on November 20, calling the situation "a truly unprecedented time."

According to the study, the annual estimate of the number of cyber incidents had been gradually increasing for almost 10 years. In 2017, the figure was about 1,500 cases, and by mid-2026, it reached approximately 4,000. Then the chart sharply rose: by November, the estimate approached 10,000 incidents.

The turning point on the chart marks the release of the GLM 5.3 weights. This is a programming model developed by the Beijing company Z.ai, formerly known as Zhipu AI. Its weights are available for download and modification, allowing the model to be used in proprietary systems and adapted for various tasks.

The U.S. National Institute of Standards and Technology (NIST) described GLM 5.3 as "the most capable open-weight cybersecurity model released to date."

Additional information on the use of AI by hackers was provided by the FBI. On November 13, the agency warned that attackers have been using the open AI framework ARTEX for cyberattacks since at least August. In some cases, they managed to gain privileged access to systems in less than three hours. However, according to the FBI, the attackers' toolkit more often included DeepSeek models rather than GLM.

How the number of attacks changed in 2026

Data from Check Point Research also indicates an acceleration in cyber activity, although the dynamics in the company's study appear less sharp. Check Point tracks the average weekly number of attacks per organization worldwide.

Month of 2026

Attacks per week per organization

Change over the year

January

2,090

+17%

February

2,086

+9.6%

March

1,995

−5%

April

2,201

+8%

May

2,055

+2%

June

2,270

+17%

July

2,336

+16%

August

2,422

+22%

September

2,803

+48%

October

2,550

+32%

In the first half of the year, growth rates remained moderate. In March, the average number of attacks was even 5% lower than in the same month of 2025. However, from July, the difference between the indicators of the two years began to increase, reaching a peak in September.

In October, activity decreased by 9% compared to the previous month, but still exceeded last year's level by 32%. At the same time, the number of ransomware attacks reached a record 1,063 cases.

One of the most active ransomware groups in October was n0n. This week, it claimed to have hacked service providers for JPMorgan Chase, Visa, and Mastercard. According to Bloomberg, the attackers gained access to customer names and their email addresses.

Why the crypto industry suffers particularly large losses

Cryptocurrency platforms remain among the most vulnerable targets and face significant financial losses. On October 29, approximately $6.3 billion was stolen in an attack on Binance. According to the publication, the incident began with a fake interview during which malware was installed on an engineer's computer. This case is described as the largest cryptocurrency theft in history.

However, major incidents occurred even before the attack on Binance. According to DefiLlama, 312 crypto project hacks were recorded in 2026, with a total damage of about $2.4 billion.

Chainalysis analysts estimate that suspected North Korean hacker groups account for nearly 89% of all losses from hacks this year. For comparison, in 2021, they accounted for about 15% of the damage.

Fraud schemes also add additional pressure on the industry. According to the FBI, U.S. residents reported losses of $11.37 billion due to cryptocurrency-related fraud in 2025. This accounts for more than half of all financial losses from cybercrime recorded by the agency.

Why artificial intelligence currently helps hackers more than defenders

Members of the Market Intelligence expert council at BeInCrypto believe that at the current stage, AI development primarily benefits attackers.

Erica Stanford, a digital assets and artificial intelligence consultant at the law firm CMS and author of the book Crypto Wars, noted that neural networks allow fraud to be carried out even by people without special skills and good language knowledge.

"This 100% helps criminals... Now anyone without skills, even if they do not have a good command of English or their native language, can very easily engage in fraud and criminal activities because AI assists them in this," Stanford stated.

Patrick Prince, co-founder of Recoveris, a company specializing in asset recovery, agreed with her. In his opinion, artificial intelligence currently expands the capabilities of criminals more noticeably than it helps law enforcement agencies.

The recovery of stolen funds also lags behind the pace of the attacks themselves. Binance managed to obtain a court order in Abu Dhabi regarding frozen funds, but most of the stolen assets remain inaccessible.

Prince expects that it will take years for law enforcement agencies to develop tools and methods of counteraction comparable to the capabilities of attackers. Until then, exchanges, banks, and their service providers will face attacks that can be executed at machine speed.

Against this backdrop, Bridgewater Associates' assessment reflects growing concerns about the economic consequences of cybercrime. However, the question remains open as to the extent to which the acceleration of attacks is specifically related to the spread of new AI models.

This material is prepared solely for informational purposes and does not constitute financial advice or a recommendation.

Comments

Checking authorization...
Loading comments...