IT Security News: Evening roundup, 2026-11-21
Summary
Several threads run through this batch: AI is showing up both as a product market and as an attack tool, and attackers keep turning trusted vendors, cloud services and identity systems into their way in. Palo Alto Networks' $4.5B move into AI model training stands out as the biggest business story. On the threat side, a North Korea-linked group hiding command-and-control on a blockchain, a Microsoft identity-service flaw, and third-party vendor breaches at banks and card networks show how supplier and platform trust is being exploited. For defenders, the common lesson is that supply-chain and AI-driven risk now reaches well beyond the organization's own perimeter.
- Palo Alto Networks agreed to buy Applied Compute in an all-stock deal valued at $4.5 billion, a sign that the security giant is moving deeper into AI model training and deployment.
- Catching Ransomware introduces the encrypts-files tool and a stricter MITRE ATT&CK T1486 detection approach, aimed at spotting ransomware that uses embedded or custom encryption rather than OS crypto APIs.
- Signal is restricting screenshots in its Android and iOS apps, a step the company frames as a response to the growing threat that AI-driven surveillance poses to private messaging.
- Cisco Talos linked the financially motivated UAT-12643 campaign to North Korean-aligned actors, who use blockchain infrastructure for command-and-control.
- JPMorgan Chase, Visa and Mastercard confirmed incidents at third-party vendors after the n0n group claimed to have leaked data, with Hackread counting over 27,000 email addresses in the posted files.
- Microsoft's November Windows 11 preview update KB5132057 adds 28 changes, including fixes for Windows Search, Windows Hello and the Run dialog.
- A software vulnerability at Belfast-based Wellington IT exposed saved payee details for members of about 40 Irish credit unions, roughly a quarter of the country's credit unions.
- An InfoWorld piece argues that enterprises are repeating familiar adoption mistakes with agentic AI, as vendors pitch decision-making agents for workflows that were already automated before.
- A report on how adversaries exploit the vulnerability ecosystem against defenders was published, though the source text available here is limited to a teaser.
- N-able says a cyberattack disrupted its Mail Assure and SpamExperts Hosted Cloud services, and the incident could have enabled credential theft.
- Suspected North Korea-aligned hackers are hiding command-and-control on the Solana blockchain, using a RAT that Cisco Talos tied to UAT-12643 and that survives Windows resets.
- The Austrian Foreign Ministry confirmed a cyberattack on its website between November 12 and an unspecified later date, exposing staff contact details, including employees at diplomatic missions abroad.
- Microsoft disclosed that attackers exploited a flaw in its Entra Connect Health service to reach administrator email addresses and tenant configuration data.
- The FBI is reportedly investigating a hack of the National Crime Information Center, and Donald Trump called the reported breach "not good."
- Hunt.io found an exposed open directory on a Netherlands VPS holding the loot of a Russian-speaking credential-theft operation that compromised 603 PCs, including bank, government and password manager accounts.
- Transport for London warned that about 265,000 Oyster Photocard customers may have had personal data exposed, after emails were sent from TfL's own system.
- Check Point argues that AI agent security needs "agentic steering" as the first response to risky behavior, since a kill switch stops both the risk and the work.
- Amazon found a Chinese-nexus actor using Cairn, an open-source AI pentesting framework, where autonomous agents found a zero-day and gained root on university targets without human direction.
- The European Commission confirmed that an attack misusing a publicly available AI tool forced several of its applications offline, with Politico breaking the story.
- BleepingComputer reports that the BADBOX malware, found on generic Android TV boxes, has returned to large-scale ad-fraud operations with roughly 40 billion daily ad requests.
Summaries written with AI from the linked source articles.
Sources in this roundup
| BleepingComputer | 3 article(s) | |
| CyberMaterial | 3 article(s) | |
| Security Affairs | 3 article(s) | |
| Security News | TechCrunch | 2 article(s) | |
| www.infosecurity-magazine.com | 2 article(s) | |
| Blog | 1 article(s) | |
| Check Point Blog | 1 article(s) | |
| Hackread – Cybersecurity News, Data Breaches, AI and More | 1 article(s) | |
| Help Net Security | 1 article(s) | |
| InfoWorld | 1 article(s) | |
| VirusTotal Blog | 1 article(s) | |
| securityweek | 1 article(s) |
Most-mentioned keywords
| data | 4 mention(s) | |
| cyberattack | 3 mention(s) | |
| agentic | 2 mention(s) | |
| aligned | 2 mention(s) | |
| blockchain | 2 mention(s) | |
| exposes | 2 mention(s) | |
| able | 1 mention(s) | |
| accounts | 1 mention(s) |
Sources
- Palo Alto Networks acquires Applied Compute for $4.5B
- Catching Ransomware: Introducing encrypts-files and Stricter MITRE ATT&CK T1486
- Signal restricts mobile screenshots as it confronts ‘existential threat’ of AI surveillance
- UAT-12643: NK-aligned actor uses blockchain C2
- JPMorgan Chase, Visa and Mastercard confirm vendor incidents after n0n leak claims
- Windows 11 KB5132057 update released with 28 changes and fixes
- Wellington IT breach exposes payee data at 40 Irish credit union
- Agentic AI repeats same old enterprise adoption cycle
- How Adversaries Are Turning the Vulnerability Ecosystem Against Defenders
- N-able Blames Service Disruption on Cyberattack
- Suspected North Korea-Aligned Group Hides C2 on Solana Blockchain
- Cyberattack on Austrian Foreign Ministry Exposes Diplomatic Staff Data
- Microsoft Entra Connect Health flaw exploited to steal tenant data
- Trump says reported FBI database hack is ‘not good’
- Hunt.io Uncovers Russian-Speaking Credential Theft Operation Targeting US Banks and Government Accounts
- TfL Warns 265,000 Photocard Customers May Have Had Personal Data Exposed
- The Kill Switch Is the Last Resort. Agentic Steering Is the First Move.
- Cairn AI Framework Used in Autonomous Zero-Day Exploitation
- European Commission confirms AI-generated cyberattack disrupted its systems
- BADBOX botnet returns with record-breaking 40 billion daily ad requests
