🚨 THE INTERNET IS GETTING HAMMERED BY CYBERATTACKS
AND NOW AI IS MAKING THE HACKERS DANGEROUSLY EFFICIENT
Look at what's happened in just days.
Zeabur suffered a breach exposing data associated with 321,810 accounts and millions of stored environment variables.
Railway warned that hackers may have accessed decrypted customer secrets.
OVHcloud had malicious software planted in its developer tools.
And Reuters uncovered something even more disturbing.
One hacker reportedly used AI agents to break into more than 100 company websites, including sites associated with Deere and SharkNinja, while doing almost none of the technical work personally.
Meanwhile, American banks are discovering unauthorized AI agents accessing customer accounts.
The really frightening shift isn't just that hackers are getting into more systems. They're beginning to automate the entire process.
A few instructions. An AI agent. Hundreds of potential targets.
The cost of launching cyberattacks is collapsing, while the damage they can cause is anything but.
Source: Axios, Reuters / Writer: Oliver
Show more
Just a few of the biggest security incidents in the past week... we're still accelerating:
* PHP2Shell: Two critical RCE bugs in PHP core
* eMudhra: Eight intermediate certificates had extremely weak private keys due to a key-generation script error that created keys with tiny values.
* GitLab: integer overflow RCE
* Plex Media Server: flaws compromised internet-facing servers used for cryptominers and proxy tools.
* Arista VeloCloud Orchestrator zero-day: Actively exploited authenticated SQL injection enabling privilege escalation.
* Node.js Brotli decoder RCE
* NGINX critical heap/use-after-free issues
* Railway breach: Attacker accessed systems storing customer environment variables and, for some, obtained decrypted contents (API keys, database credentials, private registry logins, etc.)
* Zeabur breach: ~321,810 accounts and over 3.2 million environment-variable records exposed. Company announced it is shutting down.
* LiteLLM supply-chain compromise: Malicious version 1.107.0 pushed to PyPI and Docker Hub containing a credential stealer and remote-access/backdoor capability.
* California public school data theft: 5.7M+ student records leaked, including names, dates of birth, addresses, demographics, and disability status.
* Capcom breach: Unauthorized access to ~11.77 million CAPCOM ID records
* Quest Diagnostics: Unauthorized access to Quanum Lab Services Manager system used by some clients; potential access to certain patient records.
* State Bank of India breach: customer records appearing on dark web
Show more
🚨 October 2026 was one of the most terrifying months cyberspace has ever seen.
Security researchers are already calling it "Black October."
Tens of millions of people had their personal data exposed. Critical flaws were found in software that most of the internet runs on.
Hospitals and local governments were knocked offline. And the month ended with one of the biggest crypto hacks ever.
Most of these attacks used access that was already trusted, such as a vendor, a cloud login, a certificate or an old system.
It started in the first week:
• UNIVERSITY OF ILLINOIS CHICAGO: Ransomware compromised systems at its College of Medicine.
• YAKINIKU KING: 10.79 million members' information leaked.
• DENMARK'S CPR REGISTER: 8.8 million identity records accessed without authorization.
• NIKKEI: 1.83 million records exposed, including about 150,000 with identifying information.
Then one attack hit hundreds of organizations at once:
• IDCF CLOUD: Ransomware on a single cloud provider disrupted services for 495 companies and local governments in Japan.
The leaks kept coming through mid-October:
• DAIICHI KOSHO: Around 8.7 million people's information potentially exposed.
• SKYTICKET: Around 14.64 million records reportedly affected by unauthorized access.
• BOOKOFF GROUP: Up to 6.43 million membership records compromised.
• ZEABUR: 321,810 accounts and millions of environment-variable records accessed.
• BURGER KING JAPAN: 11.18 million app account records potentially exposed.
• SOURCENEXT: 22.52 million customer records potentially exposed.
• CAPCOM: 11.77 million account records compromised.
• UNIVERSITY OF COPENHAGEN: Personal information of about 525,000 people compromised.
That's more than 100 million records in one month. Leaked names, emails and ID numbers are often used in scam emails and fake login pages.
Then critical problems were found in software that most of the internet runs on.
• EMUDHRA: On October 17, researchers found that the private keys behind eight of the Indian certificate authority's certificates were weak enough to be worked out. Browsers use these certificates to confirm a website is real. Chrome, Firefox, Apple, Android and Microsoft all blocked them within days.
eMudhra's shares fell about 20%, and the company is now leaving the global SSL certificate business.
• PHP2SHELL: On October 21, PHP released emergency fixes for two critical bugs that let attackers take over web servers without a password.
PHP runs on about 70% of web servers, including WordPress sites. Within days, the US government confirmed hackers were actively exploiting the bugs and ordered federal agencies to patch within three days.
• On the same day, attackers hijacked LiteLLM, a widely used AI developer tool, and pushed a malicious version that stole passwords, cloud keys and AI API keys. GitLab also patched a flaw rated 10 out of 10 in severity.
The PHP bugs were found with help from an AI model. AI was also behind another security problem in October: AI agents.
On October 15, India's cybersecurity agency CERT-In issued a critical warning. It said AI agents had accessed live systems at Indian organizations without authorization, including personal data and confidential business records.
Some organizations did not even detect it themselves.
US banks are seeing the same thing. JPMorgan, Morgan Stanley and Charles Schwab have started blocking outside AI agents from customer accounts.
Schwab says a single agent can try to log in 30 to 50 times a week.
In both cases, AI agents used real logins to access systems without permission.
Governments are now preparing for what happens when these systems fail.
On October 22, the UK launched its Be Ready campaign, advising people to keep at least three days of food and water at home. Similar guidance is in place across many countries:
• Germany: Up to 10 days
• Sweden: At least 7 days
• Norway: 7 days
• Estonia: 7 days
• Japan: At least 3 days, preferably a week
• France, Netherlands, Finland, Czechia: 72 hours
• Denmark: 3 days
• Canada: At least 72 hours
These plans cover cyberattacks alongside power outages, natural disasters, sabotage and military emergencies.
The biggest attacks came in the last week of October:
• University of New Mexico Health: A cyberattack disrupted hospital appointments and forced ambulance diversions.
• Daishin Securities: Personal information of 38,490 customers leaked.
• Sony's So-net: Around 2.58 million customer records affected.
• Florida Department of Education: Up to 4.9 million people potentially affected.
• CryptoBilis: Up to $93 million reportedly stolen through tampered hardware wallets.
• CoinDCX: Nearly $16 million lost through compromised legacy wallet infrastructure.
• Binance: $6.3 BILLION stolen from compromised hot wallets.
October showed how connected these risks have become.
One weak certificate can let attackers impersonate websites. One software bug can expose most of the web. One cloud provider can take down hundreds of organizations.
And one compromised wallet can put billions at risk.
Show more