Weekly AI Cybersecurity News Roundup — June 22–29, 2026

news-roundupai-securitythreat-intelligencemitre-attcknist-csfmitre-atlas

Top Stories

CISA BOD 26-04: Accelerated Patching Mandate Targets AI-Driven Exploitation

On June 10, CISA issued Binding Operational Directive (BOD) 26-04, a sweeping update to how federal agencies prioritize vulnerability remediation CISA. The new directive establishes dramatically compressed mitigation timelines: critical vulnerabilities (CVSS ≥ 9.0) must be patched within 3 days, high-severity within 30 days, and medium/low-severity within 60 days. Agencies are also empowered to defer lower-priority vulnerabilities and focus on the highest-risk items.

The directive explicitly cites AI-accelerated exploitation as a driving factor. As AI tools compress the window between disclosure and weaponization from months to days, CISA argues that traditional 30-90 day patching cycles are no longer adequate NextGov. Non-federal organizations should take note: if the US government considers 3-day patching the new baseline, private sector security teams will need similar velocity to stay ahead of AI-powered adversaries.

Check Point VPN Zero-Day (CVE-2026-50751) Under Active Exploitation Since May

Check Point disclosed a critical authentication bypass vulnerability in its VPN product line on June 8, revealing that CVE-2026-50751 had been actively exploited in the wild since May 7, 2026 Check Point Blog. The flaw exists in the deprecated IKEv1 VPN protocol and allows unauthenticated attackers to bypass VPN authentication entirely. A second related vulnerability (CVE-2026-50752) was also disclosed.

Rapid7 confirmed active exploitation with observed activity from multiple threat actors Rapid7, and ransomware affiliates were observed leveraging the flaw for initial access CyberScoop / Cybersecurity Dive. The one-month gap between first exploitation and public disclosure highlights the detection challenges organizations face when edge device compromises don’t trigger conventional alerts. Check Point has released urgent hotfixes; any organization running Quantum Security Gateway, Quantum Maestro, or Quantum Scalable Chassis with IKEv1 enabled should treat it as compromised until verified.

Sophos Uncovers AI-Powered Malware Lab Using Claude Opus for EDR Evasion

Sophos X-Ops published research on June 2 documenting a ransomware affiliate’s novel use of AI technologies to systematically evade endpoint detection and response (EDR) tools Sophos Blog. The threat actor used Claude Opus and Cursor (an AI coding IDE) to develop and iteratively refine malware that slips past multiple EDR products. The attacker established a dedicated testing environment — an AI-powered malware lab — to systematically evaluate evasion techniques against different EDR configurations.

Help Net Security described this as a “qualitative escalation in adversarial AI use,” noting that the attacker wasn’t just using AI to write code faster but to run an automated adversarial testing pipeline HelpNetSecurity. The implications are significant: if attackers can affordably rent GPU clusters to fuzz-test malware against EDR products, the traditional detection arms race shifts to one where defenders must deploy AI countermeasures at the same pace.

Breaches & Incidents

Cisco SD-WAN Hit by Two Zero-Days in Two Weeks

Cisco disclosed two actively exploited zero-day vulnerabilities in Catalyst SD-WAN Manager in a two-week span. CVE-2026-20245 (disclosed June 4) allows authenticated attackers with netadmin privileges to escalate to root on SD-WAN Manager appliances — and was observed being used to gain the highest level of access on a communications provider network CyberScoop. Cisco patched it on June 4.

Then on June 15, Cisco disclosed CVE-2026-20262, an arbitrary file write vulnerability in the same product line that PSIRT confirmed was also under limited exploitation Cisco Security Advisory. This marks the seventh actively exploited SD-WAN zero-day in 2026, according to CyberScoop. The pattern suggests that threat actors have developed systematic capability for finding and weaponizing vulnerabilities in Cisco’s SD-WAN management plane — a critical concern given SD-WAN’s role as the backbone of modern enterprise WAN architecture.

Charter/ShinyHunters Breach: 42 Million Customer Records at Risk

The ShinyHunters extortion gang breached Charter Communications via a Salesforce instance, claiming to have stolen 42 million customer records including names, email addresses, and home addresses Safestate. Charter confirmed the incident but stated that no sensitive personal information or customer proprietary network information (CPNI) was exfiltrated Have I Been Pwned. The conflicting claims leave an estimated 13 million records in dispute.

The breach follows the same playbook as the Infinite Campus and AT&T Salesforce attacks earlier this year: ShinyHunters accessing Salesforce CRM instances that handle customer data for large organizations. The pattern suggests systematic targeting of Salesforce-connected SaaS ecosystems rather than one-off opportunity hits.

Langflow CVE-2026-5027: Path Traversal RCE Under Active Exploitation

VulnCheck reported first-time exploitation of CVE-2026-5027 on June 9, a high-severity path traversal vulnerability (CVSS 8.8) in Langflow, the open-source AI development platform VulnCheck via LinkedIn. The vulnerability allows unauthenticated remote code execution on vulnerable Langflow instances SecurityWeek.

Langflow is widely used in AI/ML development workflows to build retrieval-augmented generation (RAG) pipelines and agentic AI applications. Organizations running Langflow exposed to the internet face critical risk, as the platform often holds API keys for LLM providers, database connection strings, and model configuration data. Fix: update to the latest Langflow version immediately.

AI-Powered Threats

Verizon DBIR 2026: Shadow AI Triples as Attack Surface Explodes

The Verizon 2026 Data Breach Investigations Report (DBIR), published May 19, delivered a stark statistic: 45% of employees are now regular AI users on corporate devices, up from just 15% the previous year — a tripling in twelve months Verizon. Even more concerning: 67% of those employees are accessing AI tools through personal, non-corporate accounts, bypassing enterprise security controls entirely AllCovered.

The DBIR also found that mobile social engineering success rates increased 40% Verizon DBIR 2026, suggesting that AI-generated phishing messages are becoming harder to distinguish from legitimate communications. The report’s recommendation is blunt: organizations must treat AI adoption as a data loss prevention problem, not just a productivity opportunity, and deploy controls that monitor AI tool access regardless of the account used.

Check Point: Travel Phishing Surges 122% Over Three Years

Check Point Research detailed a massive surge in travel-themed phishing attacks, with the hospitality and travel industry facing an average of 2,291 weekly attacks per organization in May 2026 — a 122% cumulative increase since 2023 Check Point. Attackers registered 47,318 travel-related domains for phishing campaigns in the first half of 2026 alone SC World.

The travel sector’s 122% growth rate Check Point Research far exceeds the 2% global average across all industries. Attackers are capitalizing on summer travel season with fake booking sites, fake airline customer support, and fake travel insurance offers — all increasingly AI-generated with near-perfect grammar and localized content.

Ransomware Surged 48% in May 2026

Check Point Research documented 698 ransomware attacks globally in May 2026, a 48% year-over-year increase compared to 472 incidents in May 2025 Check Point Blog. The surge came even as overall cyberattack volumes declined, suggesting threat actors are consolidating around high-impact ransomware operations rather than scattershot malware campaigns. Education was the most-targeted sector.

The CrowdStrike 2026 Global Threat Report noted that AI-enabled attacks surged 89% year-over-year, with mean breakout time (initial compromise to lateral movement) falling to 29 minutes CrowdStrike via LinkedIn.

Policy & Regulation

Executive Order on AI Innovation and Security

On June 2, President Trump signed an Executive Order titled “Promoting Advanced Artificial Intelligence Innovation and Security” White House. The order directs agencies to strengthen federal cyber defenses within 30 days, creates an AI cybersecurity clearinghouse, and prioritizes classified benchmarking of advanced AI models’ offensive cyber capabilities HK Law. The order also requires developers of frontier AI models to report safety test results before public deployment Skadden.

The clearinghouse is intended to serve as a central repository for threat intelligence on AI-powered attacks, enabling faster sharing across federal agencies and critical infrastructure operators. The classified benchmarking requirement mirrors China’s approach to evaluating AI models’ cyber capabilities before deployment approval.

What to Watch

  1. CISA BOD 26-04 spillover effects: Private sector organizations that contract with the federal government will likely need to adopt similar 3-day patching timelines. Start planning for compressed remediation SLAs now.

  2. Check Point VPN post-exploitation: With CVE-2026-50751 exploited since early May and credentials harvested by ransomware affiliates, expect follow-on attacks on organizations that haven’t patched or rotated IKEv1 credentials.

  3. Cisco SD-WAN targeting continues: Seven zero-days in 2026 suggests systematic reverse engineering of SD-WAN Manager. If you run Catalyst SD-WAN, treat the management plane as a critical security boundary.

  4. AI-powered EDR evasion commoditization: The Sophos findings suggest a future where AI-generated malware testing is a paid service. Watch for the first “evasion-as-a-service” offerings on criminal forums.

  5. Shadow AI policy enforcement: With Verizon’s DBIR showing 67% of AI tool usage on personal accounts AllCovered, expect major compliance-driven policy rollouts in H2 2026 — and likely the first significant fines for inadequate AI data governance.

  6. Langflow internet scanning: Following disclosure, expect Shodan/Censys scans for exposed Langflow instances. If you run Langflow, ensure it’s behind a VPN or authentication proxy.

All sources cited inline above. This roundup covers major developments from June 22–29, 2026. News is aggregated from CISA, Check Point Research, Sophos X-Ops, Verizon DBIR, Cisco PSIRT, Rapid7, SecurityWeek, HelpNetSecurity, CyberScoop, and other security news sources.

📖 Related Reads