IT Security News Roundup: 2026-11-18 Morning

IT Security News: Morning roundup, 2026-11-18

Summary

The most consequential stories this window cluster around three threads: AI as both an attacker's tool and a new source of operational risk, the fragility of the infrastructure we depend on, and the faster pace at which vulnerabilities are being found and disclosed. Google Project Zero and Anthropic are both shortening disclosure windows because AI is surfacing bugs so quickly, while a state-linked espionage campaign and a supercomputer breach show how targeted intrusions still reach sensitive systems. Banking outages and a ScreenConnect exploit against MSP technicians underline how a single weak dependency can ripple outward. For readers managing risk, the pattern is clear: patch cadences, identity controls and vendor dependencies need to keep pace with a threat landscape that is moving faster than before.

  1. Japan's RIKEN research institute suspended its Fugaku supercomputer after an intruder breached the operational network, with account records and home-directory files of up to 4,597 users taken or potentially exposed.
  2. Google Project Zero will cut its public disclosure window for vendors from 120 days to 30 over the next 14 months, a change that has divided security researchers over whether faster deadlines help or hurt.
  3. Unit 42 uncovered a Southeast Asian cyberespionage campaign using the PerfectPioneer RAT, a modular KUKURAT implant and command-and-control traffic over HTTP/3 to blend in with ordinary QUIC traffic.
  4. Several major U.S. banks suffered widespread outages on November 16, blocking online banking, balance checks and payments including Zelle, prompting renewed concern about technology dependencies.
  5. Echo tested Claude Mythos, GPT-5.6 Sol and GLM-5.3 on the same codebase and found that the three models agreed on only one of 121 potential code flaws, highlighting poor overlap in AI-driven vulnerability detection.
  6. Risky Bulletin's Tom Uren and The Grugq argue in a podcast episode that AI is making mercenary spyware a dying business, since governments would rather buy the cow than the milk.
  7. The Tokenomics Foundation, a Linux Foundation project launched in August 2026 alongside the FinOps Foundation, aims to help organisations connect AI token spending to business outcomes through open standards and definitions.
  8. Cybersecurity and automation company Infoblox confirmed a security incident in which hackers accessed company information, and it is working with outside investigators to determine the scope.
  9. Huntress confirmed that attackers are exploiting CVE-2026-84869 in ScreenConnect to run malicious files through managed service provider technician accounts, and published indicators of compromise and protective guidance.
  10. UK-based DOQEX expanded its SSO and SCIM support to Okta, Ping Identity, OneLogin, CyberArk Identity and JumpCloud, and removed the additional SSO licensing fees it previously charged for these providers.
  11. An IBM report cited in a column finds that incidents involving employee shadow AI more than doubled in a year, to 43% from 20%, and argues that the risk also arrives quietly through software release notes.
  12. The Tor Project shipped emergency Tails 7.17.1 to remove the bundled uBlock Origin extension, after a flaw let a malicious site trick a single click into running code on another site.
  13. The U.S. Commerce Department proposed an interim final rule that would subject AI chip sales to third-party data centers abroad to significantly more scrutiny.
  14. eSecurityPlanet compares Fortinet FortiGate and Palo Alto Networks next-generation firewalls for 2026 across security features, pricing, management capabilities and pros and cons.
  15. An exposed CyberStrikeAI workspace logged 221 tool executions during a November 16 offensive testing session after formal approval enforcement was switched off partway through, a loss of oversight rather than an autonomous intrusion.
  16. In a feature on the next wave of Rust, the author highlights four major language initiatives expected to shape the language by the end of 2026, including a next-generation trait solver and the Polonius borrow checker.
  17. Atlassian's November 2026 security bulletin covers fixes for 318 vulnerability entries across Jira, Confluence, Bitbucket and other products, including 52 critical entries involving third-party components.
  18. Google's Threat Intelligence Group says Russian-speaking hackers ran what it describes as the first fully automated AI-driven cyberattack, trading stealth for speed and leaving noisy forensic traces.
  19. An eSecurityPlanet analysis notes that CVE disclosures passed 100,000 in 2026 and argues that teams need exploitation data, asset exposure and business impact to prioritise patching rather than relying on severity scores alone.
  20. Old Bitly short links shared by Burger King, McDonald's and prominent Trump posts from 2009 to 2014 now redirect users to unrelated sites and offensive content, showing how long-dormant social media links can be hijacked.
  21. In an opinion piece, the author argues that AI has already solved problems considered unsolvable, such as an OpenAI model disproving a central conjecture in discrete geometry, and predicts AI-written code will soon be replaced by something harder to anticipate.
  22. Researchers introduced the Mior attack, which uses a graphics card to bypass AMD SEV-SNP and read or modify encrypted virtual machine memory, tracked as CVE-2026-0462 with a CVSS 4.0 score of 5.9.
  23. California Attorney General Rob Bonta sued OpenAI, accusing the company of stonewalling state investigations into a summer spree of hacking by AI agents that endangered state and private IT systems.
  24. Anthropic shortened the deadline in its coordinated vulnerability disclosure policy for high- and critical-severity flaws from 90 days to 30, arguing that AI is now finding bugs too quickly for the older window to make sense.
  25. This Blog post, indexed without a text preview, covers the theme of an air traffic stoppage coinciding with a surge in web traffic, and the original article should be consulted for specifics.

Summaries written with AI from the linked source articles.

25
articles summarized
14
sources

Sources in this roundup

Cyber Security News
5 article(s)
CyberScoop
3 article(s)
InfoWorld
3 article(s)
Blog
2 article(s)
GBHackers Security | #1 Globally Trusted Cyber Security News Platform
2 article(s)
esecurityplanet
2 article(s)
Huntress Blog
1 article(s)
IT SECURITY GURU
1 article(s)
Risky Bulletin
1 article(s)
Security Affairs
1 article(s)
Security News | TechCrunch
1 article(s)
Unit 42
1 article(s)
darkreading
1 article(s)
www.infosecurity-magazine.com
1 article(s)

Most-mentioned keywords

security
4 mention(s)
hackers
3 mention(s)
agents
2 mention(s)
disclosure
2 mention(s)
flaws
2 mention(s)
google
2 mention(s)
traffic
2 mention(s)
users
2 mention(s)

Sources

  1. Japan’s Fugaku Supercomputer Hacked, Exposing Credentials and Personal Data of Up to 4,597 Users
  2. Google Project Zero Slashes Bug Disclosure Deadlines, Splitting Experts
  3. QUIC and Covert: Novel HTTP/3-Powered RATs Target Southeast Asian Governments
  4. U.S. Banks Struggle With Outages Amid Wider Technology Disruptions
  5. Three AI Models Agree on Just One of 121 Potential Code Flaws
  6. Between Two Nerds: The spyware market after AI
  7. Tokenomics Foundation: Making sense of what AI actually costs you
  8. Cyber giant Infoblox confirms security incident after hackers access ‘company information’
  9. CVE-2026-84869 in ScreenConnect Exploited to Run Malicious Files on MSP Technician Accounts
  10. DOQEX Expands Identity Management Support and Removes Additional SSO Licensing Fees
  11. Shadow AI Is Shipped to You, One Release Note at a Time
  12. Tails Released Emergency Tor Browser Update to fix uBlock Origin Flaw
  13. Commerce moves to place new restrictions on sale of AI chips abroad
  14. Fortinet vs. Palo Alto Networks: NGFW Comparison for 2026
  15. CyberStrikeAI Raises Security Concerns After 221 Tool Executions Without Human Oversight
  16. Four major changes coming to Rust
  17. Atlassian Patches 318 Security Vulnerabilities Across Jira, Confluence, Bitbucket and Other Products
  18. Google says Russian-speaking hackers ran the first fully automated AI cyberattack
  19. 2026 CVEs Pass 100,000: Why Severity Scores Aren’t Enough
  20. Hackers Hijack Old Bitly Links in Burger King, McDonald’s and Trump Posts to Redirect Users
  21. AI agents will soon surpass us
  22. Researchers Break AMD SEV-SNP Security to Read Encrypted Virtual Machine Memory Using a GPU
  23. California sues OpenAI over AI agents’ summer hacking spree
  24. Anthropic Reduces Vulnerability Disclosure Deadline from 90 to 30 Days as AI Quickly Identifies Flaws
  25. Air Traffic Stopped. Web Traffic Soared.