Jun 16's Top Cyber News NOW! - Ep 1154

At a glance

Eight stories spanning AI policy, phishing at scale, supply chain attacks, and espionage. The through-line: defenders need MFA, detection logging, and access controls—basics that nation-state actors still exploit because they're not implemented.

Stories covered

Is Anthropic's AI ban hurting defenders more than attackers?

What happened: The US administration forced Anthropic to take its advanced Mythos and Fable AI models offline over jailbreak concerns. Cybersecurity leaders from Adobe, Zoom, and Sophos are now urging the government to reverse the restrictions, arguing that blue teams need these tools to detect AI-powered threats.

Why it matters: Threat actors are already using AI to automate kill chains and find vulnerabilities faster. Banning advanced models from defenders while bad actors access competitors' versions creates an asymmetric disadvantage for defenders.

What to do: Monitor this policy debate. Assume the genie stays out of the bottle—tools like this won't stay banned long, and you'll need to learn how to defend against AI-assisted attacks regardless.

---

How did a China-based phishing service steal $1.9 billion?

What happened: The FBI, Google, and Black Lotus Labs dismantled Outsider Enterprise, a phishing-as-a-service operation that used AI-powered kits and over a million fraudulent URLs to impersonate carriers like AT&T and Verizon. The scheme resulted in 3.8 million stolen credit card records and nearly $2 billion in losses.

Why it matters: This is enterprise-grade crime: 9,000 fake websites, a million malicious URLs, and AI automating the text message spam. This is proof that phishing scales when you layer AI on top of traditional social engineering.

What to do: Train users on vishing and phishing at scale. Implement DMARC, SPF, DKIM. For SMS, educate users that carriers never ask for verification via unsolicited text. Monitor passive DNS for lookalike domains; Canary tokens help detect when users click.

---

Why did One Password pay $250M for an Israeli identity startup?

What happened: One Password acquired Aono, an Israeli company that automates access permissions for cloud infrastructure and AI agents. The deal signals that managing non-human identities—service accounts, API keys, tokens for AI systems—is now a critical security gap.

Why it matters: As organizations deploy AI agents, those agents need credentials, permissions, and audit trails. Managing access for thousands of non-human identities is a new attack surface; threat actors stealing agent credentials gain lateral movement at scale.

What to do: Begin inventorying service accounts, API keys, and agent credentials in your environment. Plan for identity and access management (IAM) refresh to cover non-human identities. Expect this to be a hot topic at RSA 2027.

---

How did attackers hide backdoors in WordPress plugins for months?

What happened: Threat actors compromised Awesome Motive, the vendor behind Optin Monster, Trust Pulse, and Push Engage, and planted malicious JavaScript that created hidden admin accounts and installed backdoor plugins. The attack potentially exposed 1.2 million WordPress sites before it was detected and patched on June 12.

Why it matters: This is supply chain attack at the developer level, not post-build. Attackers integrated malicious code into the source before release, bypassing many detection methods because the "legitimate" vendor signed it.

What to do: Immediately audit any WordPress sites running Awesome Motive plugins for unfamiliar admin accounts. Update to patched versions (post-June 12). Implement alerts on admin account creation in WordPress. If critical, restore from pre-June 12 backups and review access logs for lateral movement.

---

How did China spy on US medical research for a year undetected?

What happened: A Chinese state-sponsored group (UNC6508) infiltrated a US medical university connected to military research, deployed custom malware to steal credentials from the REDCap research platform, and remained undetected for over a year. They stole credentials and used them to access domain admin accounts.

Why it matters: Nation-state operators are patient. They steal creds and sit on them; they escalate to domain admin (crown jewels) and harvest sensitive research on medical, public health, and defense technologies.

What to do: Enforce phishing-resistant MFA (FIDO2) on all accounts, especially domain admins and research platforms. Limit domain admin use to schema changes only; use Just-In-Time admin activation. Monitor for credential theft via EDR and passive DNS. If you run REDCap, audit logs immediately.

---

Can Microsoft Copilot leak emails and MFA codes with a single click?

What happened: Researchers at Veronus found a critical flaw in Microsoft 365 Copilot that combined prompt injection with URL parameter manipulation to trick Copilot into exfiltrating emails, files, calendar data, and MFA codes. Microsoft has patched it; no known exploitation in the wild.

Why it matters: This is a new attack surface: AI systems connected to internal data via old web vulnerabilities. Prompt injection + URL parameter injection = data exfiltration through "trusted" Microsoft infrastructure, hard for WAFs to detect.

What to do: Monitor for unusual Copilot activity (email forwarding, calendar access) in your audit logs. Be aware that prompt injection attacks can bypass traditional web controls. Review and limit Copilot's permissions to necessary data sources only.

---

How many organizations has the Shiny Hunters group hacked via PeopleSoft?

What happened: Shiny Hunters exploited a PeopleSoft zero-day to breach over 100 organizations, including the University of Nottingham and now the Council of Europe, stealing payroll, HR, banking, and medical records. The group now claims it stole 297 GB from the Council of Europe.

Why it matters: When a single vulnerability affects widely deployed software, threat actors go on a feeding spree. Shiny Hunters is systematically compromising PeopleSoft instances and collecting data while the window is open.

What to do: If you run Oracle PeopleSoft, apply all available patches immediately. Audit access logs for unusual activity. Assume breach if you haven't patched; begin incident response. Monitor dark web and paste sites for your organization's name and employee data.

---

Why is a $10M ransom demand on a port authority a warning?

What happened: Annabis ransomware group breached Italy's Adriatic Port Authority via phishing, stole contracts, employee records, and port security plans, and demanded $10 million. The port says only 2% of data was actually compromised and refused to pay.

Why it matters: Maritime infrastructure is critical national security. Ransomware on ports disrupts global supply chains. The attack chain: phishing → unpatched Sonic Wall VPN, Solar Winds, Cisco SSL VPN → no MFA → full compromise.

What to do: If you manage critical infrastructure (ports, utilities, energy), mandate MFA on all internet-facing VPN and remote access systems. Patch Sonic Wall, Citrix, Cisco, and Solar Winds immediately. Assume phishing will succeed; MFA stops the next step.

---

Key takeaways

• MFA is non-negotiable. Nation-state actors, ransomware gangs, and phishing services all exploit the absence of MFA to escalate from initial access to domain admin. Stop treating it as optional.
• Supply chain attacks now strike at developer commit time. Awesome Motive's compromised source code infected 1.2M downstream sites. Assume third-party code is a vector; audit dependencies and monitor for anomalous behavior post-deployment.
• AI amplifies both attack and defense. Phishing-as-a-service with AI scales to a million URLs; Copilot prompt injection leaks internal data. Defenders need AI tools, detection logging, and access controls—and they need them now.
• Domain admin accounts are still crown jewels—and still misused. The Chinese espionage group exploited stolen domain admin creds for a year. Limit DA use, enforce JIT admin activation, and alert on every DA account creation.
• Canary tokens and detection-focused NIST CSF controls (Detect, Respond, Recover) are under-invested. Most programs mature "Identify and Protect" and neglect "Detect." Admin account creation, email forwarding, and Copilot data access must be alertable.

Topics covered

anthropic, AI policy, phishing as a service, outsider enterprise, credit card theft, one password, aono, identity access management, AI agents, wordpress, awesome motive, supply chain attack, backdoor, redcap, china espionage, domain admin, MFA, microsoft 365 copilot, prompt injection, peoplesoft zero day, ransomware, adriatic port authority, sonic wall VPN

Show notes generated from the live transcript using AI on Tue, 16 Jun 2026 18:43:40 GMT. Errors? Open the YouTube replay for the source of truth.