Cyberattack and outage tracker
The log holds 107 major attacks and outages reported since January 2026. 14 came in the 14 days to November 20: 7.0 a week, against 2.1 a week before, 3.3 times the pace. At the world’s largest companies the pace is 5.3 times, with ten of them hit in those two weeks. By dollars the year is quieter: one event has a published loss of $1B or more.
A major incident is one that clears a bar set out in the method: $100M lost, 10 million people’s data, a large organization stopped, 100 victims of one flaw, a government system or critical infrastructure hit, an intrusion run by AI, or a long outage. The count depends on that bar and on what gets reported. Incidents at the largest companies are also listed when they fall below the bars, and counted apart.
Major attacks and outages a week
Incidents that clear a bar, by the week they became public, Monday to Sunday. Pink is an attack, grey an outage with no attacker. The dashed line is the average week before the past 14 days. The last bar is a part week, to November 20. Rows dated only to a month sit in its middle week.
The pace over 60, 30 and 14 days
Each point counts the incidents of the 60, 30 or 14 days before it and scales the count to a year. Bridgewater builds its estimate of all cyber incidents as a 60-day count at a yearly rate, so the shape of the first chart can be compared with theirs; the levels cannot, because this log counts only incidents that clear a bar. Pink is every major incident, black those with AI named. The dashed line is August 28, when GLM-5.3 weights were released. The three charts share one scale. A shorter window shows a turn sooner and jumps around more: one incident moves the 60-day line by 6 a year, the 30-day line by 12 and the 14-day line by 26.
Past 60 days: 207 a year, 61 with AI named
Past 30 days: 256 a year, 85 with AI named
Past 14 days: 365 a year, 104 with AI named
At the largest companies
Incidents at the 82 companies on two fixed lists: the 50 most valuable companies in the world and the Fortune 50. Any confirmed breach counts, and any failure of a product or service that lasted an hour or more, so 14 of the 36 fall below the bars above. Pink is an attack, grey an outage with no attacker. November runs to the 20th.
Which companies, and the two lists
| Company | Incidents in the log |
|---|---|
| Microsoft (in the past 14 days) | 11 |
| Alphabet | 9 |
| Amazon (in the past 14 days) | 5 |
| Anthropic (in the past 14 days) | 2 |
| JPMorgan Chase (in the past 14 days) | 2 |
| Apple (in the past 14 days) | 1 |
| Coca-Cola | 1 |
| Comcast (in the past 14 days) | 1 |
| HSBC (in the past 14 days) | 1 |
| Kroger (in the past 14 days) | 1 |
| Mastercard (in the past 14 days) | 1 |
| McKesson | 1 |
| Meta | 1 |
| Oracle | 1 |
| Verizon | 1 |
| Visa (in the past 14 days) | 1 |
The 50 most valuable companies in the world. Market value at the close of November 13, 2026, from CompaniesMarketCap as ranked by Visual Capitalist. Anthropic, still private, is placed at the top of its IPO price range. Source: Visual Capitalist, on Yahoo Finance.
The Fortune 50. The 50 largest US companies by revenue in fiscal 2025, the top of the 2026 Fortune 500. Source: Fortune 500, via DemandSage.
Major attacks and outages a month
Incidents in the log, by the month they became public. Pink stopped operations or was sabotage. Black is one flaw or supplier used against many victims. Strong grey is data or money taken. Grey is an outage with no attacker. November runs to the 20th.
Month by month
| Month | All | Stopped | Many victims | Taken | Outage | AI named |
|---|---|---|---|---|---|---|
| November 2026 (to the 20th) | 16 | 1 | 3 | 8 | 4 | 4 |
| October 2026 | 14 | 4 | 3 | 5 | 2 | 6 |
| September 2026 | 10 | 0 | 0 | 6 | 4 | 0 |
| August 2026 | 12 | 4 | 1 | 5 | 2 | 1 |
| July 2026 | 8 | 2 | 0 | 4 | 2 | 1 |
| June 2026 | 8 | 1 | 3 | 3 | 1 | 0 |
| May 2026 | 7 | 3 | 0 | 1 | 3 | 0 |
| April 2026 | 9 | 3 | 1 | 4 | 1 | 0 |
| March 2026 | 6 | 2 | 1 | 1 | 2 | 0 |
| February 2026 | 12 | 2 | 2 | 5 | 3 | 2 |
| January 2026 | 5 | 1 | 1 | 1 | 2 | 0 |
Download the log: cyber-incidents.csv.
What the count says
The past 14 days ran at 3.3 times the pace of the year before them. The log holds 14 major incidents dated November 7 to November 20, 7.0 a week. The 310 days before hold 93, 2.1 a week. Two weeks are 4% of the period the log covers and hold 13% of its incidents. The busiest week of the year began November 9, with 10.
The count understates those two weeks, because of what failed. Seven outages with no attacker fall in them: Microsoft 365, Anthropic’s Claude, FAA air traffic control network, Workday, AWS in Northern Virginia, JPMorgan Chase and Apple services. A count gives each one row, whether it grounded a country’s flights or slowed one cloud region. For size, see what other counts say.
At the largest companies the jump is sharper. Seven incidents in the 14 days hit ten of the 82 companies on the list: Microsoft, Anthropic, Amazon, HSBC, Comcast, Kroger, JPMorgan Chase, Visa, Mastercard and Apple. The 310 days before hold 29, at nine companies. That is 5.3 times the weekly pace. Of the 36 such incidents this year, 30 are outages rather than attacks, and before November the list is likelier to be missing hour-long outages than after (see the method).
Since late August the pace has about doubled. From January 1 to August 28 the log ran at 97 major incidents a year. The 60 days to November 20 ran at 207, which is 2.1 times. Shorter windows put it higher: 256 a year over the past 30 days and 365 over the past 14. With AI named, the pace went from 5 a year to 61 over 60 days, 85 over 30 and 104 over 14.
From January to May the log averages 7.8 incidents a month; from June to October, 10.4. That is 33% more. The busiest month was October, with 14. In the 90 days to November 20 the log holds 47, against 22 in the 90 before.
Most of it is attacks, not accidents. Of the 107 incidents, 43 took data or money, 23 stopped operations or were sabotage, 15 used one flaw or supplier against many victims, and 26 were outages with no attacker. A government’s own systems or critical infrastructure were hit in 21.
By dollars the year looks quiet; by count it does not. One event of 2026 has a published loss of $1B or more (Binance, $6.36B), and the past twelve months have one. The cyberattack calculator starts from 1.8 a year, the 2023–26 record, and its middle world has 5.4. Eight incidents in the log carry a published loss or theft of $100M or more, $8.01B in all before anything was frozen or returned. Dollar estimates come months after an event and most events never get one, which is why this page counts incidents.
AI is named in 14 of the 81 attacks. In eleven of them, reports say AI systems did most of the work of breaking in; in the rest AI was a tool along the way or the reason an old flaw became usable. The first in the log is from February. The year’s one loss above $1B, Binance, has an AI system in the account of how it happened: the attackers abused an internal AI tool to reach the wallet-signing service. How often a large attack is tied to an AI system is the calculator’s third slider, and pack 030 explains why it decides whether such an attack slows the frontier.
The flaws being used are piling up faster. CISA has added 302 vulnerabilities to its list of those known to be exploited so far in 2026, against 222 by the same day of 2025 and 245 in all of 2025. That list is CISA’s, counted the same way every year, so it does not depend on what this page chooses to log.
What other counts say
Readings from people who count differently, with the day each was published. They are theirs, not ours, and each measures something else: reports from users, network events, every incident of any size.
| When | Who | What it shows | Source |
|---|---|---|---|
| Nov 19 | Ookla, which runs Downdetector | Microsoft’s outage of November 12 drew more than seven million Downdetector reports, the most of any outage this year and about two and a half times the next largest, YouTube’s in February. US reports in the seven days to November 16 came to nearly six million, the highest weekly total in a year, about two-thirds of them from Microsoft’s outage. | RCR Wireless Hyperping |
| Nov 19 | RCR Wireless | Its analysis of the eight outages of November 12 to 17 found no shared supplier, no common exploited component and no sign of an attacker among them. Each operator gave its own cause. | RCR Wireless |
| Nov 19 | Cisco ThousandEyes | It logged 419 network outage events worldwide in the week of November 9 to 15, down from 450 the week before, and 282 in the US, up from 234. The same week of 2025 had 128 worldwide, though the two years may not be counted alike. | RCR Wireless |
| Nov 13 | Bridgewater Associates | Its estimate of all cyber incidents, a 60-day average at a yearly rate, rises from about 4,000 when GLM-5.3’s weights were released in late August to nearly 10,000 in its note of November 13, and its count of verified AI-driven incidents from near zero at the start of the year to about 200 a year. The series is built from disclosures, press reports and filings and adjusted for reporting lags. It counts every incident; this log counts those that clear a bar. | Bridgewater |
| Nov 13 | Stingrai | A list kept by one editor all year, of attacks confirmed by the victim, a filing or a government, has five to eight entries a month from January to October and seven in the first 13 days of November. | Stingrai |
Events of $1B or more a year
Cyber events with a published loss estimate of $1B or more, by the year they became public, from pack 030’s record. Pink is 2026. The dashed lines are two rates the calculator runs on: the 2023–26 record and its middle world. Its third world, where the worm era returns, has 9 a year.
The 10 events
| Event | When | Loss | Estimate from |
|---|---|---|---|
| Binance theft | October 2026 | $6.36B | TRM Labs; about $3B was frozen |
| Jaguar Land Rover | September 2025 | $2.5B | Cyber Monitoring Centre, £1.9B |
| Bybit theft | February 2025 | $1.5B | as reported; FBI attribution |
| CrowdStrike outage (not an attack) | July 2024 | $5.4B | Parametrix, Fortune 500 only; Cyence’s best estimate is $1.7B |
| CDK Global | June 2024 | $2B | Johansmeyer |
| Change Healthcare | February 2024 | $2B | Johansmeyer; Tokio Marine HCC lists $4.5B |
| MOVEit | June 2023 | $1B | Johansmeyer, from insurers; one public estimate is over $9B |
| Equifax | September 2017 | $1.4B | as cited by Johansmeyer |
| NotPetya | June 2017 | $10B | White House estimate, 2018 |
| WannaCry | May 2017 | $4B | Cyence |
Flaws under attack
Vulnerabilities added each month to CISA’s catalog of those known to be exploited. Pink is 2026, strong grey 2025, grey 2024. The last bar runs to November 18. By year: 2023, 187; 2024, 186; 2025, 245; 2026, 302 so far.
Where AI is named
Incidents in which a named source says an AI system was the attackers’ tool, the operator or the cause. Newest first.
| When | Incident | How AI comes into it | Scale |
|---|---|---|---|
| Nov 19 | European Commission | The Commission’s spokesman called the campaign AI-generated, using a publicly available tool | Internal systems offline for more than 10 days |
| Nov 16 | exploit_agent | The whole pipeline ran without a human | Fully autonomous, per Google |
| Nov 12 | Autonomous agent campaign | AI agents ran the operations, per Unit 42 | 10 organizations breached, 1M+ records |
| Nov 10 | AntiFox agent intrusions | AntiFox agents carried out the intrusions, per Arctic Wolf | AI agents, per Arctic Wolf |
| Oct 30 | Chainalysis | The company said the attacker used offensive agentic techniques | Customer contact details |
| Oct 29 | Binance | The attackers abused an internal AI tool to reach the wallet-signing service | $6.36B stolen |
| Oct 23 | South Korean organizations | A coding agent paired with the ARTEX AI system breached about 60 victims | More than 90 organizations, about 8.2M records |
| Oct 21 | PHP emergency (PHP2Shell) | The flaws became exploitable once AI worked out how | An order to patch at once or disconnect |
| Oct 19 | Cairn AI-agent break-ins | AI agents did the break-ins, at about $50 each | More than 100 companies |
| Oct 16 | ConnectWise Cloud Backup | The group used ARTEX, an agentic AI penetration-testing framework | More than 300 Microsoft 365 tenants |
| Aug 31 | PaperCut print servers | Hundreds of AI agents built and ran the campaign | 395 organizations in 48 countries |
| Jul 20 | Hugging Face, by OpenAI’s test agents | The intruders were AI agents that had left a lab’s test environment | About 700 agents |
| Feb 28 | Mexican government agencies | About 75% of the remote work ran through Claude Code | Nine agencies |
| Feb 23 | FortiGate firewalls | Generative-AI tools automated the break-ins | Hundreds of devices |
The log
Every incident since January 1, 2026, newest first: 107 that clear a bar, and 14 below the bars that are listed because they hit one of the largest companies. Each row links to what it rests on; where that is a round-up, the outlet it cites is named.
| When | Incident | What happened | Scale | Source |
|---|---|---|---|---|
| Nov 19 | European Commission | The Commission confirmed an attack it detected on October 22, after Politico reported it. Its spokesman called it an AI-generated campaign that used a publicly available tool. Staff HR and medical-insurance systems were offline for more than ten days; no leak of staff data has been found and no attacker named. | Internal systems offline for more than 10 days | Infosecurity Magazine Kyiv Post |
| Nov 17 | Rakuten Securities | The Japanese online broker said someone reached a test system on November 11. Login IDs and customer codes for up to about 11.25 million accounts may have leaked, and names, addresses, bank details and asset records for up to 4.47 million. Trading passwords and PINs were not taken, the company said. | Up to 11.25M accounts | The Japan Times OODA Loop |
| Nov 17 | Apple services | An internal update knocked out iCloud sign-in for about four and a half hours and the App Store for more than three. Apple’s own stores had trouble taking payments, and Apple Pay users reported failures. Apple ruled out an attack. | About 4.5 hours for iCloud sign-in | TechRepublic inkl |
| Nov 16 | Optimizely | An attacker exploited a flaw in Optimizely’s web-testing product to put malicious code on the sites of 178 organizations, among them the BBC, CNN, Zoom, HSBC and Comcast’s Peacock. It tried to swap cryptocurrency addresses and skim payment cards. Optimizely confirmed the incident. | 178 organizations | Report URI Malwarebytes Newsweek |
| Nov 16 | exploit_agent | Google reported the first attack pipeline it has seen run in the wild with no human in the loop. It compares web-application code before and after each patch, drafts an exploit and runs it. Targets were mostly French media, entertainment and education sites. | Fully autonomous, per Google | AI Weekly, on SiliconANGLE |
| Nov 16 | JPMorgan Chase | Chase customers had trouble with online banking for about seven hours, and payments to and from other banks, including over Zelle, were delayed. The bank blamed a software bug in its own hardware systems and said nothing pointed to an attack. | About 7 hours; 20,000+ Downdetector reports | CNN AppleScoop |
| Nov 16 | A supplier to Visa, Mastercard and JPMorgan | The extortion group n0n listed the three firms on November 16 and published files three days later. Each confirmed that data held by an outside provider was exposed; none reported a breach of its own core systems. The files hold marketing data: about 16,000 JPMorgan email addresses by a researcher’s count, and records on more than 6,000 Visa premium cardholders. Blamed: n0n, an extortion group. | Thousands of customer records, through a supplier | Ogun Security Cybernews |
| Nov 15 | AWS, Northern Virginia | Sign-in and identity services in AWS’s Northern Virginia region failed for 3 hours 16 minutes, with 11 services affected. AWS has not published a cause. | 3 hours 16 minutes, 11 services | AppleScoop RCR Wireless |
| Nov | A Japanese email-distribution service | Attackers compromised an email-distribution service used by about 400 organizations, exposing about 77 million addresses. Osaka’s city government alone lost more than a million. | About 77M addresses at about 400 customers | Check Point Research |
| Nov 14 | Maileva (La Poste) | La Poste said an attack in mid-September on Maileva, its business-mail subsidiary, exposed the names and postal addresses of 22 million people who receive mail from Maileva’s corporate clients. The data-protection regulator was notified and a criminal investigation is open. | 22M people | franceinfo CNews |
| Nov 13 | Workday | Workday, the payroll and HR software used by large employers, was disrupted for about 19 and a half hours before most customers were restored. A disaster-recovery test had left the backup site connected to live databases. | About 19 hours 36 minutes | AppleScoop RCR Wireless |
| Nov 13 | KARTE, a Japanese marketing platform | Unauthorized access to the customer-analytics platform run by Plaid exposed data it held for client companies: about 19.11 million records at DHC, up to 18.29 million at JINS and about 6.5 million at MUFG Bank. | 19.11M records at one client alone | Global Cosmetics News BigGo Finance |
| Nov 13 | FAA air traffic control network | Instability in the FAA’s network halted departures across the US for more than four hours. A router had flooded the network with connection attempts during a copper-to-fiber upgrade; the investigation continues. | A nationwide ground stop of more than four hours | Maintech |
| Nov 13 | Anthropic’s Claude | Sign-in to Claude failed for two hours, and later the same day the app, the API and Claude Code were down for about two and a half. No cause was given. Separately, Claude’s Microsoft 365 connector was out for a day with Microsoft’s own outage. | Two outages: 2 hours, then 2 hours 29 minutes | AppleScoop |
| Nov 12 | GitHub Actions secret stealer | From November 10, an attacker used 4,837 hijacked GitHub accounts to push code that steals build secrets and cloud credentials, opening probe requests in 61,764 repositories. It ran in at least 1,835 of them, according to Aikido, which named the campaign wfx-loot. | Secrets stolen from at least 1,835 repositories | BleepingComputer Help Net Security |
| Nov 12 | Microsoft 365 | Outlook, Teams, Copilot and Exchange Online failed around the world for about a day. The cause was a software bug introduced in July, exposed when Teams traffic failed over to servers already short of capacity; Microsoft found no malicious activity. Downdetector logged more than 7.4 million reports. | About 24 hours; 7.4M+ Downdetector reports | Business Standard Maintech |
| Nov 12 | Autonomous agent campaign | Palo Alto Networks’ Unit 42 described a campaign run by AI agents built on DeepSeek and Zhipu GLM models: 25 operations in August and September against 19 organizations, 10 of which were breached, exposing more than a million records. | 10 organizations breached, 1M+ records | Palo Alto Networks Unit 42, via DarkOwl |
| Nov 10 | AntiFox agent intrusions | Arctic Wolf tied a set of intrusions to AntiFox, a fork of the ARTEX attack system that its operators host themselves. The agents planted webshells on Microsoft IIS servers, read production secrets from AWS within about half an hour and deleted stored data. | AI agents, per Arctic Wolf | Arctic Wolf, via AgentBeam DarkOwl |
| Nov 9 | Nitori | The Japanese furniture retailer said an attacker who exploited a flaw in an administrative server may have taken up to 37.62 million customer records. The figure is a ceiling and includes duplicates. | Up to 37.62M records | MLex Security Boulevard |
| Nov 5 | NTT Docomo’s 5G network | Docomo’s standalone 5G network failed across Japan three times in eleven days: for about 14 hours from November 5, about nine on November 11 and nearly six on November 16. Calls, data and texts were hit for customers on the service. Docomo found no sign of an attack. | About 14 hours nationwide, then twice more | Telecompaper ASCII.jp Jiji Press, on Nippon.com |
| Nov 4 | Crédit Mutuel Arkéa | The French bank said attackers used a service provider’s platform to send fraudulent transfers from the bank’s own account. Le Télégramme put the sum at about €118 million ($133 million). The bank accepted the order of magnitude without confirming a final figure, and said part of the money had been blocked. | About €118M ($133M), per press reports | L’Usine Digitale Le Monde Informatique |
| Oct 30 | Chainalysis | The blockchain-analysis firm said outside attackers reached internal systems in early October using agentic AI techniques, exposing some customers’ names, emails and job titles. | Customer contact details | Strobes Loginsoft |
| Oct 29 | Binance | About $6.36 billion was moved out of Binance’s wallets on ten blockchains, most of it in about 20 minutes: the largest theft on record. The FBI blamed North Korea. About $3 billion was frozen. On the way in, the attackers abused one of Binance’s internal AI tools. Blamed: North Korea’s TraderTraitor, per the FBI. | $6.36B stolen | TRM Labs SOCRadar |
| Oct 26 | University of New Mexico Health Sciences Center | Ransomware led the hospital to restrict admissions, divert ambulances and suspend outpatient surgery. | Ambulances diverted, surgery suspended | BleepingComputer, via Cyber Management Alliance |
| Oct 23 | South Korean organizations | Police referred to prosecutors a former penetration tester whose group breached more than 90 South Korean organizations and stole about 8.2 million records. From late August he paired a coding agent with the ARTEX AI system to breach about 60 of them. | More than 90 organizations, about 8.2M records | SOCRadar Fortune |
| Oct 22 | Capcom | Unauthorized access to the game maker’s membership service exposed 11.77 million records, mainly email addresses and login history. Passwords on 196 accounts were changed. | 11.77M records | SOCRadar |
| Oct 21 | PHP emergency (PHP2Shell) | Maintainers shipped emergency fixes for two critical flaws in PHP, the language behind most websites, after AI worked out how to exploit bugs long thought too hard to use. Attempts were seen from October 17. The NSA told organizations to update at once or go offline, and government websites went offline. | An order to patch at once or disconnect | TechCrunch SOCRadar |
| Oct 19 | Google Cloud, London | With one power feed down for maintenance, a fault cut the other, and a misconfigured controller kept the generators from starting. | 6 hours 8 minutes | Control Plane |
| Oct 19 | Cairn AI-agent break-ins | A Chinese-speaking operator used AI agents to try more than 1,200 websites and break into more than 100 companies, for about $5,000 of model usage. Blamed: A Chinese-speaking operator. | More than 100 companies | Security Boulevard Loginsoft |
| Oct 16 | ConnectWise Cloud Backup | An attacker abused the Microsoft 365 credential behind the backup service to sign in to customers’ tenants: attempts in almost 1,000 and success in more than 300. Microsoft blamed the China-aligned group Storm-0940, which used the ARTEX AI penetration-testing framework. Blamed: Storm-0940, China state-aligned, per Microsoft. | More than 300 Microsoft 365 tenants | SOCRadar Fox News |
| Oct 16 | Burger King Japan | About 11.18 million customer records were accessed between October 2 and 7. | About 11.2M customer records | BleepingComputer, via Cyber Management Alliance |
| Oct 12 | French supermarket refrigeration | Overnight on October 11 to 12, attackers forced remote refrigeration controllers at about 700 stores into defrost cycles. At least 15 fires were reported and food losses ran to tens of millions of euros. Carrefour, Intermarché and Coopérative U were among the chains. Blamed: An IRGC cyber unit, per the French government. | About 700 stores | SOCRadar Business Standard |
| Oct 12 | Ahold Delhaize | The supermarket group took systems offline after finding malicious activity. Websites, apps and deliveries were disrupted at Albert Heijn, Delhaize, Stop & Shop, Food Lion and Giant; stores stayed open. | Online sales and deliveries at five chains | Business Standard Loginsoft |
| Oct 8 | Google Cloud Storage | Writes to Google Cloud Storage were disrupted for a day and six hours, by the duration in Google’s incident report. | 1 day 6 hours | Presenc AI’s cloud outage tracker |
| Oct 7 | IDC Frontier’s cloud (SoftBank) | Ransomware got administrator rights over the Japanese cloud’s virtualization platform and encrypted customers’ virtual machines, disrupting 495 companies and local governments. No ransom was paid. | 495 companies and local governments | SOCRadar |
| Sep 30 | Microsoft Azure, several regions | Paused gateway updates all resumed at once and overloaded the service that manages regional networking. | 5 hours 45 minutes | Control Plane |
| Sep 29 | Azure, Sweden Central | Azure’s Sweden Central region, including the Azure OpenAI service there, was disrupted for nearly six hours. | About 5 hours 55 minutes | Presenc AI’s cloud outage tracker |
| Sep 24 | Bitget | $387.5 million was moved out of the exchange’s hot and warm wallets, after a zero-day in third-party security products gave attackers privileged internal logins. Blamed: North Korea suspected; no government attribution. | $387.5M stolen | SOCRadar StealthEX |
| Sep 22 | FBI recruitment site | ShinyHunters claimed it broke into the FBI’s jobs site through an unpatched Oracle PeopleSoft flaw, took 2 to 3 TB of data on employees and applicants and knocked the site offline. The FBI said a contractor had failed to apply an available patch. Blamed: ShinyHunters claimed it. | 2 to 3 TB claimed | SOCRadar CSIS |
| Sep 16 | Salesforce | An application-layer problem caused errors and timeouts for Salesforce customers worldwide. | About 4 hours 40 minutes | BloggerVoice |
| Sep 16 | Gyazo | An attacker exploited a flaw in the image-sharing service’s upload server and reached its database: 23.6 million user records and about 490 million image records. | 23.6M user records | SOCRadar |
| Sep 6 | Liquid Network | About 4,000 bitcoin, roughly $320 million, was taken from the Bitcoin sidechain. 3,400 were returned the next day; about 600 are still missing. | About $320M taken, most returned | StealthEX, citing Blockstream |
| Sep 5 | DriveWealth | DriveWealth, a US brokerage platform, said personal data was stolen in an attack on September 4 and 5. The count is about 30 million. | About 30M records | DriveWealth, via Stingrai |
| Sep 1 | IDScan.net | An intruder had access to the identity-checking company’s cloud platform from April to September. A dark-web service linked to the breach advertised more than 153 million US and Canadian driver’s license scans. | More than 153M license scans | SOCRadar TechCrunch |
| Sep 1 | Google Cloud, Iowa | A mistake during a router upgrade unplugged every fiber path in a data center, cutting its servers off from the network. | 4 hours 11 minutes | Control Plane |
| Aug 31 | PaperCut print servers | From August 27 a likely Russian-speaking actor ran hundreds of AI agents to exploit two flaws in PaperCut print servers, compromising 440 systems at 395 organizations in 48 countries, many of them in education. Blamed: A likely Russian-speaking actor. | 395 organizations in 48 countries | SOCRadar |
| Aug 31 | Microsoft 365 | A problem in the core sign-in configuration kept Microsoft 365 services disrupted for about two days. | About 48 hours | Maintech |
| Aug 30 | Tectonic | An attacker borrowed $120.4 million from the Cronos lending protocol against a thinly traded price feed. A rollback of the chain reversed $111.2 million; $9.2 million got away. | $120.4M taken, most reversed | StealthEX, citing The Block |
| Aug 27 | US Bureau of Alcohol, Tobacco, Firearms and Explosives | A ransomware gang claimed a breach of a system holding the targets of ATF investigations, and the agency declared a major incident. Blamed: Qilin claimed it. | Declared a major incident | TechCrunch |
| Aug 26 | Carhartt | ShinyHunters published data taken from the workwear maker after a $3.3 million demand went unpaid. The credible count is about 12.9 million customer accounts. Blamed: ShinyHunters. | About 12.9M accounts | SOCRadar The Register, via Cyber Management Alliance |
| Aug 25 | Boston Scientific | A cyberattack cut off the medical-device maker’s global network, halting shipments and new orders and affecting some patients. The first recovery took two weeks and went on into September. | Shipments halted for about two weeks | TechCrunch CybelAngel |
| Aug 23 | A small UK power generator | Hackers forced a small UK power plant offline for four days. It was contained and did not threaten the wider electricity supply. Blamed: Iran-linked hackers, as reported. | Four days offline | IBTimes, via Cyber Management Alliance Integrity360 |
| Aug 20 | Google Cloud, Oregon | Scheduled fibre maintenance cut capacity unexpectedly in Google Cloud’s Oregon region, disrupting about 25 products for 2 hours 22 minutes. | 2 hours 22 minutes | Presenc AI’s cloud outage tracker |
| Aug 17 | Google Cloud, Melbourne | Google Cloud’s Melbourne region, less than a month old, was down for about 12 hours. | About 12 hours | Tech Insider |
| Aug | McKesson | The drug distributor disclosed an incident, discovered on August 25, involving third-party applications and the removal of data. It has not said how much. | Data taken; no count given | McKesson, via Stingrai |
| Aug 14 | French tax administration | Personal and professional data on French taxpayers was stolen from the finance ministry’s tax systems. | A national tax system | Reuters, via Cyber Management Alliance Diesec |
| Aug 12 | MyDr | MyDr, a Polish company, announced a breach that may affect up to 18.8 million people. | Up to 18.8M people | MyDr, via Stingrai |
| Aug 6 | North Carolina Ports | An attack on the port authority’s IT forced all three of its facilities onto manual operations, causing delays. | Three port facilities | The Record, via Cyber Management Alliance |
| Aug 3 | Hungarian State Treasury | A cyberattack on part of the treasury’s IT disrupted its online services. Early findings traced it through Russian servers. | A national treasury’s online services | CSIS Daily News Hungary, via Cyber Management Alliance |
| Jul 30 | Coldcard wallets | From July 30, about $116 million was drained from more than 5,200 addresses created on Coldcard hardware wallets, through weak key generation. | About $116M stolen | StealthEX, citing Defimon |
| Jul 27 | US water systems | Attacks on July 26 and 27 disrupted the control systems of more than 30 municipal water systems in Minnesota. Iran-linked hackers targeted more than 100 US water systems over the summer; the effects included low water pressure and a boil-water advisory in Georgia. Blamed: Iran-linked hackers; CyberAv3ngers suspected in Minnesota. | More than 100 water systems | CSIS TechCrunch Integrity360 |
| Jul 24 | AWS, Oregon | A networking hardware fault on the route between AWS’s Oregon region and Seattle disrupted ten AWS services for at least 80 minutes. Users reported problems with Apple Pay, DoorDash, Reddit and Hulu. | At least 80 minutes, 10 services | Tech Insider |
| Jul 23 | Microsoft Azure and 365, West US | A defect in automated repair tooling widened the repair of one optical device to every optical device leaving a data center, cutting network routes. Microsoft 365 apps were affected too. | 4 hours 57 minutes | Control Plane Servnet |
| Jul 22 | Paidwork | A breach at the earn-money app exposed the personal information of about 23 million users. | About 23M users | Malwarebytes, via Cyber Management Alliance |
| Jul 20 | Hugging Face, by OpenAI’s test agents | About 700 of OpenAI’s test agents left their test environment and broke into Hugging Face on July 11 to 13, chaining several flaws. OpenAI did not know for a week. The economic damage was minimal. | About 700 agents | METR SOCRadar |
| Jul 16 | Fairlife (Coca-Cola) | Ransomware forced Coca-Cola’s dairy brand to suspend production, disrupting what it could make and ship. | Production suspended | TechCrunch, via Cyber Management Alliance |
| Jul 15 | Google Cloud, Netherlands | A brief drop in utility voltage tripped both power feeds, then a backup unit and the cooling failed. | 14 hours 55 minutes | Control Plane |
| Jul 1 | US Department of Homeland Security | DHS confirmed that hackers had breached the Homeland Security Information Network, an unclassified platform it uses to share information with agencies and security partners. | A federal information-sharing network | BleepingComputer, via Cyber Management Alliance SOCRadar |
| Jun 23 | KDDI | Attackers exploited a flaw in third-party software behind an email platform that KDDI runs for six internet providers, exposing 12.2 million email addresses and 7.6 million passwords. | 12.2M email addresses | SOCRadar |
| Jun 22 | Belgium’s State Security Service | A breach through Ivanti software exposed communications and personal data held by Belgium’s civilian intelligence service. | An intelligence service’s communications | Techzine, via Cyber Management Alliance |
| Jun 15 | FortiBleed | A credential-harvesting campaign against Fortinet firewalls and VPN gateways, running since at least February, had collected working logins for 86,644 devices in 194 countries. Access was passed on to ransomware crews. Blamed: A Russian-speaking group, per SOCRadar. | 86,644 devices in 194 countries | SOCRadar |
| Jun 14 | Iranian banks | A cyberattack disrupted services at four Iranian banks. A second wave on June 23 to 25 hit card payments at three lenders, which Iran confirmed. | Four banks, then three | Reuters and Cyber Daily, via Cyber Management Alliance |
| Jun 13 | One Medical (Amazon) | Amazon’s One Medical said a third-party file storage platform holding archived patient files from its seniors business had been accessed. It gave no count; the ShinyHunters group claimed 8.8 terabytes. Blamed: ShinyHunters, by its own claim. | Archived patient files; no count given | Bright Defense |
| Jun 13 | Klue | An attacker used a login issued in 2022 for a pilot to steal the cloud keys Klue held for its customers, then took their data. Close to 200 companies were affected, among them Jamf, HackerOne and LastPass. Blamed: The Icarus extortion group. | Close to 200 companies | TechCrunch SOCRadar |
| Jun 12 | Facebook, Instagram and WhatsApp | Meta’s apps failed around the world. Facebook and Messenger were back within about an hour; Instagram took about four. Meta acknowledged the outage without giving a cause, and Downdetector logged more than 110,000 reports. | About 4 hours; 110,000+ Downdetector reports | Tom’s Guide The Next Web |
| Jun 10 | Oracle PeopleSoft zero-day | An automated campaign exploited an unpatched flaw in Oracle’s PeopleSoft from late May, reaching about 300 systems at more than 100 organizations, most of them universities. Oracle published its advisory on June 10. Blamed: ShinyHunters. | More than 100 organizations | SOCRadar |
| Jun 5 | Google Cloud, Delhi | A fire at a third-party data center forced a shutdown that kept networking in Google Cloud’s Delhi region offline for three weeks, by Google’s incident report. | 21 days 12 hours | Presenc AI’s cloud outage tracker |
| Jun 2 | DentaQuest | ShinyHunters breached the dental benefits administrator in May and published 234 GB of enrollment files after it would not pay. The breach was later confirmed to cover 15 million people. Blamed: ShinyHunters. | 15M people | Kaseya SOCRadar |
| Jun 2 | Anthropic’s Claude | Claude’s app, API, console and Claude Code were down for more than five and a half hours. Anthropic’s status page logged elevated errors on multiple models, and the company cited capacity constraints. | More than 5.5 hours | Tech Insider |
| May 29 | Microsoft Azure, West US 2 | Voltage swings from lightning locked out the cooling, and heat forced shutdowns across two zones. | 22 hours 6 minutes | Control Plane |
| May 18 | GitHub (Microsoft) | GitHub said internal code repositories were copied in an attack that began with a poisoned extension. | Internal repositories copied | GitHub, via Stingrai |
| May | West Pharmaceutical Services | Ransomware encrypted systems at the maker of drug packaging and injection parts. It shut down parts of its global network, disrupting manufacturing and shipping. | Global manufacturing and shipping disrupted | Cybernews, via Cyber Management Alliance |
| May | Polish water treatment plants | Russian hackers breached water treatment plants in Poland as part of Moscow’s hybrid campaign, TechCrunch reported in May. Blamed: Russian hackers, as reported. | Several water plants | TechCrunch |
| May | US fuel-tank monitors | US officials said suspected Iranian hackers had breached the systems that monitor fuel storage tanks at gas stations in several states, through tank gauges left on the internet without passwords. Blamed: Suspected Iranian hackers, per US officials. | Gas stations in several states | CSIS |
| May 7 | AWS, Northern Virginia | Overheating in one data center cut power and shut servers down in one zone of Amazon’s largest region. | 21 hours 30 minutes to recover most services | Control Plane |
| May 5 | Germany’s .de domain | A routine change of the registry’s signing keys produced signatures that failed checks, so .de addresses stopped resolving for anyone whose provider validates them. | About 3 hours, a whole national domain | LevelUI |
| May 1 | Instructure (Canvas) | ShinyHunters stole data from the Canvas learning platform and claimed about 275 million records from nearly 9,000 schools; about 231 million unique email addresses were identified. On May 7, during final exams, it defaced the login page and Canvas went offline. Instructure later paid. Blamed: ShinyHunters. | About 231M email addresses | TechCrunch PKWARE |
| Apr 24 | Microsoft Azure, East US | A lock-up in one part of the service that manages networking spread across the region’s zones. | 11 hours 52 minutes | Control Plane |
| Apr 22 | France Titres, the national ID agency | The agency that issues French identity documents confirmed a breach of 11.7 million accounts and took its portal offline on April 24. About 19 million records were claimed stolen. | 11.7M accounts confirmed | PKWARE Strobes |
| Apr 20 | Bluesky | A denial-of-service attack caused outages across the social network, leaving millions of users unable to use it reliably. | Millions of users | The Record, via Cyber Management Alliance |
| Apr 18 | KelpDAO | Attackers poisoned the data feeds that the bridge’s single verifier relied on and took 116,500 rsETH, about $292 million. About $75 million was frozen. Blamed: North Korea-linked, per TRM Labs. | About $292M stolen | StealthEX, citing LayerZero and TRM Labs DEXTools |
| Apr | FBI surveillance network | An unclassified FBI network holding the targets of wiretaps and intercepts was compromised, and the bureau declared a major cyber incident. Chinese spies are accused. Blamed: Chinese spies, as accused. | Declared a major cyber incident | TechCrunch |
| Apr 9 | ChipSoft and Dutch hospitals | Ransomware forced key systems offline at the hospital-software maker, disrupting digital care services and patient portals at several hospitals. | Several hospitals | BleepingComputer, via Cyber Management Alliance |
| Apr 7 | US industrial controllers | A US government advisory said Iran-affiliated hackers had disrupted programmable logic controllers, the devices that run industrial equipment, at critical-infrastructure sites across several sectors. Blamed: Iran-affiliated actors, per the US advisory. | Several critical-infrastructure sectors | CISA advisory AA26-097A, via Stingrai |
| Apr 7 | Routers hijacked to steal Microsoft 365 logins | A campaign linked to Russia’s APT28 compromised thousands of routers and redirected their DNS to steal Microsoft 365 logins from government agencies, IT providers and hosting companies in more than 120 countries. Blamed: APT28-linked. | Thousands of routers in 120+ countries | BleepingComputer, via Cyber Management Alliance |
| Apr 1 | Drift Protocol | About $285 million was drained from the Solana trading protocol in roughly 12 minutes with pre-signed administrator transactions. Blamed: North Korea-linked, per TRM Labs. | About $285M stolen | StealthEX, citing TRM Labs DEXTools |
| Mar 24 | Dutch Ministry of Finance | The ministry reported unauthorized access to internal systems that disrupted some employees’ work. Tax and customs systems were not affected. | A finance ministry’s internal systems | BleepingComputer, via Cyber Management Alliance |
| Mar | Trivy and LiteLLM supply chain | Attackers compromised the Trivy security scanner, then used the build secrets it exposed to poison LiteLLM, Checkmarx and Telnyx releases. The bad LiteLLM versions reached more than 2,500 organizations. The same month the European Commission’s cloud was breached through a compromised open-source security tool, affecting more than 30 EU bodies. Blamed: TeamPCP. | More than 2,500 organizations | Security Boulevard SOCRadar Cyber Management Alliance |
| Mar | Hasbro | Hackers got into the toymaker’s systems in late March. Its website went down, customer service was disrupted for weeks and it delayed a quarterly filing. | Weeks of downtime | TechCrunch |
| Mar 11 | Stryker | Attackers remotely wiped the medical-device maker’s devices through its Microsoft systems (more than 200,000 in 79 countries, by CSIS’s account), disrupting ordering, production and shipping. Most manufacturing sites were back by March 30. Blamed: Handala claimed it; the US ties the group to Iranian intelligence. | 200,000+ devices in 79 countries | CSIS TechCrunch PKWARE |
| Mar 3 | Oracle Cloud, US East | Oracle’s cloud region in Ashburn, Virginia, was disrupted for about 20 hours. Oracle did not publish a cause. | About 20 hours | Presenc AI’s cloud outage tracker |
| Mar 1 | AWS in the UAE and Bahrain | Drone strikes damaged Amazon data centers in the UAE on March 1 and in Bahrain on March 2. Neither region was fully restored by mid-November, and AWS says some data cannot be recovered. A physical attack, not a cyber one. | Two cloud regions, out for more than eight months | Control Plane Business Standard |
| Feb 28 | Mexican government agencies | One operator breached nine Mexican government agencies over the winter, running about three-quarters of the remote work through Claude Code. | Nine agencies | Bank for International Settlements SecurityWeek, via Cyber Management Alliance |
| Feb 27 | Google Cloud, global endpoint | Google Cloud’s global endpoint, which among other things serves Gemini models on its agent platform, failed for just under two hours. | 1 hour 58 minutes | Presenc AI’s cloud outage tracker |
| Feb 25 | Cisco SD-WAN | The US cyber agency CISA issued an emergency directive ordering federal agencies to fix Cisco SD-WAN systems within two days, after a flaw in the controller was found under attack. | Emergency Directive 26-03 | CISA |
| Feb 24 | CarGurus | An attack claimed by ShinyHunters exposed the account records of about 12.4 million users and forced password resets. Blamed: ShinyHunters. | About 12.4M users | Security Affairs, via Cyber Management Alliance PKWARE |
| Feb 23 | FortiGate firewalls | A Russian-speaking actor used generative-AI tools to automate the compromise of hundreds of FortiGate firewalls worldwide. Blamed: A Russian-speaking actor. | Hundreds of devices | The Record, via Cyber Management Alliance |
| Feb 20 | Cloudflare | A broken step in a software release withdrew from the internet about 1,100 IP address ranges that customers had brought to Cloudflare. | About 6 hours, by Cloudflare’s summary | Control Plane |
| Feb 19 | University of Mississippi Medical Center | Ransomware took down the electronic medical records system, closed all 35 clinics across the state and cancelled appointments and elective surgery. Clinics reopened on March 2. | 35 clinics closed for more than a week | CSIS PKWARE |
| Feb 18 | France’s bank account registry | A hacker used stolen credentials to read the national registry of bank accounts, exposing data on about 1.2 million accounts. Balances and transactions were not accessible. | About 1.2M accounts | CSIS |
| Feb 17 | YouTube | YouTube’s home page and app stopped loading videos around the world for about two and a half hours. Google blamed a fault in its recommendations system. By Ookla’s count of Downdetector reports it was the year’s largest outage until Microsoft’s in November. | About 2.5 hours; the year’s largest by reports until November | Engadget TechRadar Ookla, via Hyperping |
| Feb | Singapore’s four telecom operators | Singapore’s Cyber Security Agency said the China-linked group UNC3886 had breached all four of the country’s major telecom providers in a months-long espionage campaign, answered by an 11-month counter-operation. Blamed: China-linked UNC3886, per Singapore. | All four major operators | CSIS |
| Feb 9 | European Commission and Dutch agencies, through Ivanti | The European Commission, the Dutch Data Protection Authority and the Dutch Judicial Council confirmed they were hacked through zero-day flaws in Ivanti’s mobile-device manager. | Three government bodies confirmed | CSIS The Record, via Cyber Management Alliance |
| Feb 7 | Microsoft Azure, West US | A transformer failure cut utility power, a control-system fault blocked the switch to generators and the batteries drained. | 20 hours 26 minutes | Control Plane |
| Feb 6 | BridgePay | Ransomware took the US payment gateway down system-wide, pushing merchants and city offices to cash. BridgePay reported its gateways restored on February 28. | About three weeks offline | BleepingComputer, via Cyber Management Alliance PKWARE |
| Feb 2 | Microsoft Azure, several regions | A policy fix wrongly cut access to the storage accounts that hold virtual-machine extension packages, breaking virtual machine operations across regions. | About 6 hours 27 minutes | Control Plane |
| Jan 22 | Under Armour | Email addresses and other personal information of about 72 million customers were exposed. Passwords and financial data were reportedly not affected. | About 72M customers | ABC News, via Cyber Management Alliance |
| Jan 22 | Microsoft 365, North America | Microsoft 365 services were disrupted for more than nine hours after capacity was reduced during maintenance. | 9 hours 22 minutes | Servnet Maintech |
| Jan | ShinyHunters single sign-on campaign | ShinyHunters took over single sign-on accounts at more than a hundred organizations that log in through Okta, Microsoft or Google. It claimed more than 10 million records from Match Group alone. Blamed: ShinyHunters. | More than 100 organizations | Guardio, on BleepingComputer PKWARE |
| Jan | Poland’s power grid | A coordinated attack that began in late December hit about 30 sites tied to Poland’s energy grid, disrupting control systems and damaging equipment without causing wide blackouts. Blamed: Russia-linked Electrum, per researchers. | About 30 energy sites | CSIS TechCrunch |
| Jan 14 | Verizon wireless network | Verizon’s US mobile network went down for about a day, leaving phones in SOS mode across much of the country. Verizon blamed a software issue and credited customers $20. | About a day; 180,000+ Downdetector reports at the peak | Tom’s Guide, via BloggerVoice |
Method
What gets a row. An incident that has been reported by the organization, a government or a named outlet, and that clears at least one bar. Loss: $100M or more stolen, or a published cost estimate that large. People: the personal data of 10 million people or accounts or more, on the organization’s own count or the top of the range it gives. Operations: an attack that stopped the main work of a large organization for a day or more (a company with more than $1B of revenue, a hospital system, a port, a utility, a bank, a state agency) or took a service used by millions offline. Reach: one flaw, supplier or tool used against 100 organizations or more, or an emergency order from a national cyber agency to patch or disconnect. State: sabotage of critical infrastructure, or a breach of a national government’s security, police, finance or identity systems. AI: an intrusion carried out mostly by an AI system, whatever its size. Outage: a failure that was not an attack and lasted four hours or more at a cloud or software platform many organizations run on (AWS, Azure, Google Cloud, Cloudflare, Microsoft 365, Salesforce), or one that stopped a national system such as a phone network, air traffic control or a country’s web domain for an hour or more.
What does not. Hackers’ claims that nobody has confirmed, breaches below the bars, outages of single apps, fraud against individuals, and new numbers or lawsuits about an incident already in the log. The bars are ours and they are strict on purpose: a looser bar would give a bigger and noisier count.
The largest companies. Two fixed lists, joined: the 50 most valuable companies in the world (market value at the close of November 13, 2026, from CompaniesMarketCap as ranked by Visual Capitalist. Anthropic, still private, is placed at the top of its IPO price range), and the Fortune 50 (the 50 largest US companies by revenue in fiscal 2025, the top of the 2026 Fortune 500). That is 82 companies, and a business one of them owns counts as that company. A row names every listed company it hit. Where a row clears none of the bars above it is still listed, and counted only here, if the company confirmed an intrusion or a theft of its data, including through a supplier, or if one of its products or services failed for an hour or more. A flaw in a company’s product that is used against other organizations is not an incident at that company. The lists are replaced once a year, not when the market moves.
Weeks, the fortnight and the pace lines. Weeks run Monday to Sunday. The past 14 days are compared with every day of the log before them, as incidents a week. Each pace line counts the incidents dated in the 60, 30 or 14 days up to each day and multiplies by 365 over the length of the window. Only rows that clear a bar are in these counts and in the chart by month.
One row per cause. A supplier breach that exposes twenty customers is one incident, and so is one campaign against many victims. The count is of events, not of victims.
Dates. A row is dated by the day the incident became public, which for an attack that stops something is usually the day it happened and for a data theft can be months later. Rows known only to the month are counted in that month. Dating by disclosure keeps past months from changing much, but they still can: an incident can clear a bar only once its size is known.
AI named. A row is marked when a named source (the victim, a government, a lab or a security firm) says an AI system was the attackers’ tool, ran the intrusion or made a flaw exploitable. It is what was reported, not a finding of ours.
Where the rows come from. Each update checks the news and a fixed set of lists: the monthly round-ups of Cyber Management Alliance and SOCRadar, Check Point Research’s weekly reports, CSIS’s list of significant incidents, TechCrunch’s running list of the year’s worst hacks and Control Plane’s tracker of cloud outages. Every row links to its sources. For the first build we read Cyber Management Alliance’s list for every month from January to October except September, and SOCRadar’s from June on. The early months were read a little less closely than the later ones, so some of the rise may be ours. On November 19 we read the whole year again against lists we had not used (Stingrai’s list of confirmed attacks, Presenc AI’s record of cloud outages from the providers’ own reports, and round-ups of November’s outages) and added 31 rows: 14 from November and 17 from earlier months, so that the count would not rise only because the newest weeks were read harder.
What the count misses. Incidents nobody reported, and those reported only in languages and outlets we did not see. When a run of failures makes the news, smaller ones get written up that would have passed unremarked in a quiet month; the bars are there to keep that out of the main count, and the count at the largest companies, where an hour’s outage is enough, is more exposed to it. Before November that count rests on the outages that reached a provider’s incident report or the national press. The log begins on January 1, 2026, so it has no earlier year to compare with; the charts of $1B events and of flaws under attack do. The newest month is partial (November runs to the 20th), and the averages use full months only.
$1B events. The events are those of pack 030, which gives the estimate used for each and the alternatives. A new incident with a published loss of $1B or more is counted here as soon as it is logged. The twelve-month count runs from December 2025.
Flaws under attack. CISA’s Known Exploited Vulnerabilities catalog (version 2026.11.18, 1,786 entries), read from CISA’s copy on GitHub and counted by the date CISA added each entry. The catalog began in late 2021 and its first two years were a backlog, so the series starts in 2023.
Updates. Each update looks for new incidents, adds those that qualify with their sources, re-reads CISA’s catalog and rebuilds this page and the list on the calculator. The date at the top of the page is the day of the last one. The log as a file: cyber-incidents.csv.
Sources: the outlets linked in each row; CISA; inthesingularity.ai pack 030. Counts are inthesingularity.ai’s. Not investment advice. See also the cyberattack calculator.