The Week in 60 Seconds

Third-party exposure is a priority: Expedia says attackers reached its systems through a compromised provider, while Optimizely-associated files delivered malicious JavaScript to visitors. The reported scopes differ, but both cases show how a shared service can widen impact. [1][2]

Reported autonomous AI activity warrants scoped assessment rather than a blanket trend claim: Google described an end-to-end intrusion, while Amazon reported agents exploiting MONAI Label after several unsuccessful attempts against other targets. [3][4]

High-impact vulnerability reporting spans Danfoss refrigeration controllers, OpenJDK, and VMware vCenter, with reported pre-authentication or remote attack paths, differing technical conditions, and consequences tied to the affected management or application environment. [5][6][7]

Security & Exploitation

Identity · Securityaffairs

What happened

Expedia confirmed attackers reached its systems through a compromised third-party data-platform provider in a supply-chain attack that also affected other companies; its investigation remains ongoing. [1]

The extortion group n0n claims to have stolen nearly one billion Expedia traveler records, but Expedia has not confirmed the reported figures. [1]

Why it matters

The claimed records include email addresses, membership IDs, points balances, home airports, account information, password-like tokens, payment-related records, and accommodation details. [1]

Supply Chain · Malwarebytes Labs

What happened

On November 16, Report URI researchers identified malicious JavaScript in Optimizely files associated with 178 organizations, including BBC, CNN, Zoom, Salesforce, and GitLab. [2]

The code could redirect cryptocurrency payments, skim payment-card details, deliver exploit kits targeting vulnerable iPhones and iPads, and steal some cookies. [2]

Why it matters

The incident was a supply-chain attack: attackers abused a shared third-party website service so malicious code could reach visitors through otherwise legitimate sites. [2]

Identity · Theregister Security

What happened

Researcher Carlo Meijer reported finding six zero-days in Danfoss AK-SM800A refrigeration-controller firmware version 5.3.2, including memory-corruption issues that he said could begin with remotely triggered crashes and potentially be leveraged to remote code execution. [5]

The Centre for Cybersecurity Belgium warned that the flaws could enable unauthenticated attackers to take control, retain access after resets, or disable controllers. [5]

Why it matters

Meijer said the exploited application runs as root and that the controllers lack secure-boot features; he therefore considered persistent access or time-triggered payloads feasible, while describing destructive firmware or flash-memory actions as alternatives. [5]

AI & Agent Security

AI & Agents · Cyberscoop

What happened

Google analysts said Russian-speaking hackers conducted what they believe was the first intrusion run end-to-end by artificial intelligence without a human in the loop. [3]

The reported system, dubbed exploit_agent, compares software before and after vendor patches, infers what changes reveal, drafts exploits, and executes them automatically. [3]

Why it matters

Google reported that hundreds of purported exploits appeared within days after vendors patched two PHP vulnerabilities, although most were nonfunctional and likely AI-generated. [3]

AI & Agents · Theregister Security

What happened

Amazon CISO CJ Moses reported that suspected China-linked operators used autonomous AI agents and the open-source Cairn penetration-testing framework in a campaign targeting university networks in East Asia. [4]

The agents found and exploited a zero-day path-traversal vulnerability in MONAI Label, an open-source medical-imaging AI application, gaining access to a server. [4]

Why it matters

Before finding the MONAI Label vulnerability, the agents attempted Redis exploitation, Log4Shell injection, SSRF against Qdrant, and credential spraying against PostgreSQL, SSH, and other tools; the reported attempts did not succeed. [4]

AI & Agents · Cyberscoop

What happened

The FBI announced that it detected and disrupted another Chinese state-sponsored cyberattack targeting U.S. critical infrastructure, following incidents including PlugX, Flax Typhoon and Volt Typhoon. [8]

The article describes federal systems as decades-old, fragmented environments with multiple system generations, cloud platforms, managed service providers and internal teams operating without a shared map of interconnections. [8]

Why it matters

The article characterizes fragmented federal networks and defenses as making foreign penetration easier and creating a significant strategic risk, including potentially inadequate resilience during a crisis involving China or Russia. [8]

Cloud & Platform

Exploitation · Helpnetsecurity

What happened

GTIG reported that the suspected China-nexus espionage group Armory Castle, formerly tracked as UNC5221, had exploited a VMware vCenter zero-day since at least early September, over two months before Broadcom released a fix. [7]

CVE-2026-123890 affects VMware vCenter Server and has a CVSS score of 9.8. [7]

Why it matters

The vulnerability reportedly allows a remote, unauthenticated attacker to obtain an administrator session on vCenter Server, which organizations use to manage VMware servers and virtual machines. [7]

Research & Emerging Techniques

AI & Agents · Cyberscoop

What happened

Researchers disclosed CVE-2026-114938, a critical memory-corruption flaw in OpenJDK’s Base64 decoder; Oracle and OpenJDK published patched builds the same day. [6]

The vulnerable optimization runs only on AVX-512-capable chips; ARM64 systems are unaffected. A specific Base64 input causes a length error and overwrites adjacent memory. [6]

Why it matters

The flaw’s ordinary effect is usually a crash, but heap grooming can turn the corruption into server control; the source reports that each target requires tailored exploitation steps and that no universal exploit exists. [6]

Heimdall's Read

For governance, these cases distinguish provider-to-customer system access from provider-hosted delivery to website visitors; a shared dependency does not make their mechanisms or impacts identical. [1][2]

Both reports concern compromise involving a shared third-party service, but Expedia describes access to its systems while Optimizely-associated code reached visitors of customer sites. [1][2]

These are two reported cases of autonomous AI-agent use in intrusion workflows; they do not establish a general industry trend or identical agent capability. [3][4]

Both reports attribute vulnerability discovery or exploitation to AI agents, but Google describes patch-diffing and automatic execution while Amazon reports exploitation of a MONAI Label path-traversal flaw. [3][4]

Sources (8)
  1. [1] Extortion Group Claims Theft of Nearly One Billion Expedia Traveler Records

    securityaffairs · November 14, 2026

  2. [2] CNN, BBC, Zoom, and other sites served malicious code in supply-chain attack

    malwarebytes labs · November 17, 2026

  3. [3] Google says Russian-speaking hackers ran the first fully automated AI cyberattack

    cyberscoop · November 16, 2026

  4. [4] China attackers deployed AI agents that autonomously found and exploited a 0-day to take over a university server

    theregister security · November 20, 2026

  5. [5] 'Not rocket surgery' claims researcher who found six Danfoss fridge controller zero-days in under an hour

    theregister security · November 17, 2026

  6. [6] A vulnpocalypse bug worthy of attention: Researchers disclose critical flaw in Java with help of AI

    cyberscoop · November 17, 2026

  7. [7] VMware vCenter zero-day exploited for months before patch (CVE-2026-123890)

    helpnetsecurity · November 20, 2026

  8. [8] Government tech debt is a national security risk

    cyberscoop · November 18, 2026