The number of cyber incidents worldwide has more than doubled over the past three months. The key trigger was the public release of a Chinese AI model with advanced hacking capabilities — attacker-grade tools are now available to any user without special training.
Analysts at Bridgewater Associates are recording a tectonic shift in the digital threat landscape. The firm is already building into its models a one-in-three probability of an economically dangerous cyberattack within a year. Banks, payment services, and crypto exchanges — the segments where liquidity and sensitive data are concentrated — are most at risk.
What Triggered the Sharp Rise
Bridgewater's annualized estimate grew steadily for nearly ten years — from roughly 1,500 incidents in 2017 to about 4,000 by mid-2026. Then the chart went almost vertical and by November approached 10,000 incidents. One event is noted at the turning point — the public release of weights for GLM 5.3 in late August.
GLM 5.3 is a code-oriented AI model from Beijing-based Z.ai (formerly Zhipu AI) that anyone can download and modify. The US National Institute of Standards and Technology (NIST) called it in its review "the most cyber-specialized of existing public models."
The FBI warned that cybercriminals have been using ARTEX — an open-source AI-based hacking platform — since at least August. In some cases, attackers gained privileged access in less than three hours. At the same time, the use of DeepSeek models, rather than GLM, was most frequently recorded.
Monthly Cyberattack Growth Rates
Check Point Research data points in the same direction, although the rise is less sharp. The company tracks the average number of weekly attacks per organization worldwide: in January — 2,090 (+17% year over year), in February — 2,086 (+9.6%), in March — 1,995 (−5%), in April — 2,201 (+8%), in May — 2,055 (+2%), in June — 2,270 (+17%), in July — 2,336 (+16%), in August — 2,422 (+22%), in September — 2,803 (+48%), in October — 2,550 (+32%).
In the first half of the year, growth was moderate, and in March the figure even fell below the 2025 level. Starting in July, the gap compared with the previous year widened every month and peaked in September. In October, the number of attacks declined 9% from September but remained 32% above last year's value. That month, the number of ransomware attacks reached a record 1,063.
One of the most active ransomware groups in October was n0n. It claimed an attack on suppliers working with JPMorgan Chase, Visa, and Mastercard. The attackers gained access to customers' names and email addresses.
Why the Crypto Industry Suffers the Largest Losses
Crypto-related platforms are losing the most. Binance lost about $6.3 billion on October 29 — the largest theft in cryptocurrency history. The breach began with a fake job interview: the attackers installed a virus on an engineer's computer.
Before that case, DefiLlama had already recorded 312 hacks of crypto projects totaling $2.4 billion in 2026. According to Chainalysis estimates, almost 89% of losses from attacks this year are linked to suspected North Korean groups. In 2021, their share was 15%.
Additional losses are borne by fraud victims. In 2025, US citizens reported $11.37 billion in losses from crypto-related fraudulent operations — more than half of all FBI-registered cybercrimes involving financial losses.
Defense Is Already Losing the Race
AI more often ends up on the side of the attackers. Now a person without skills and even without knowledge of English or their native language can easily commit fraud and crimes, because AI helps with this. Right now, AI gives more capabilities to criminals than it helps the police.
Recovering stolen funds also lags far behind. For example, Binance obtained a court ruling in Abu Dhabi to freeze assets, but most of the stolen funds remain inaccessible. It will take several years before law enforcement can mount a comparable response. Until that happens, every exchange, bank, and trading platform faces attackers operating at machine speed, and the trend shown by the Bridgewater curve is not changing.
Cryptalist analysis: The release of open cyber-oriented models has created an irreversible asymmetric shift: the cost of attack has fallen, while the cost of defense has remained high. For the crypto sector, this means a structural increase in the risk premium and growth in security spending as a share of operating budgets. Institutional players should build scenarios with more frequent incidents into their risk management models rather than treating them as one-off events.