May 18, 2026 · Policy, Regulation & AI Industry Developments
POLICY & REGULATION
Tags: Alert | United States | Security
Date: May 14, 2026
CISA adds Cisco Catalyst SD-WAN flaw to KEV catalog with three-day federal remediation deadline
The U.S. Cybersecurity and Infrastructure Security Agency added CVE-2026-20182, a critical authentication bypass vulnerability in Cisco Catalyst SD-WAN Controller and SD-WAN Manager, to its Known Exploited Vulnerabilities catalog on May 14, 2026. The KEV listing carries a remediation deadline of May 17, 2026 for Federal Civilian Executive Branch agencies under Binding Operational Directive 22-01. Cisco confirmed the flaw, which carries a CVSS score of 10.0, is being actively exploited in limited attacks attributed by Cisco Talos with high confidence to UAT-8616, the same threat cluster behind earlier exploitation of CVE-2026-20127.
The vulnerability allows an unauthenticated remote attacker to bypass peering authentication and obtain administrative privileges on affected SD-WAN Controllers, formerly marketed as vSmart, and SD-WAN Managers, formerly vManage. A successful exploit permits attacker login as an internal high-privileged non-root user account, which can then be weaponized to access NETCONF and manipulate network configuration across the SD-WAN fabric. CISA issued Emergency Directive 26-03 and published Hunt and Hardening Guidance alongside the KEV listing.
Federal agencies must complete remediation by the May 17 deadline; private-sector organizations operating Catalyst SD-WAN should treat the KEV listing as authoritative evidence of active exploitation and prioritize patching accordingly. Organizations should also review SD-WAN management plane access controls and audit administrative session activity for indicators of pre-patch compromise, including unauthorized SSH key additions, NETCONF configuration changes, and privilege escalation attempts consistent with documented UAT-8616 post-exploitation behavior. KEV listings establish a federal compliance obligation under BOD 22-01 and a defensible private-sector standard of care under emerging state cybersecurity statutes.
Tags: Guidance | United States
Date: May 13, 2026
Department of Education’s AI priority for grant programs takes effect
The U.S. Department of Education’s Final Priority and Definitions on Advancing Artificial Intelligence in Education took effect on May 13, 2026, supplementing the Secretary’s existing supplemental priorities on Evidence-Based Literacy, Educational Choice, and Returning Education to the States. The priority, published April 13, 2026 in the Federal Register as FR Doc. 2026-07087 under Docket ED-2025-OS-0118, allows the Secretary to apply AI-related criteria to discretionary grant competitions across the Department’s programs, providing federal funding leverage for AI integration in K-12 and postsecondary education.
The priority establishes definitions and evaluation criteria the Department may apply to grant applicants pursuing AI-enabled instructional tools, AI literacy programs, embedding AI and computer science into teacher preparation, and AI workforce readiness initiatives. The Secretary retains discretion to use the entire priority for a grant program or to apply individual components selectively, and the priority is codified under 34 CFR Part 75. The action follows the December 2025 Executive Order on Ensuring a National Policy Framework for Artificial Intelligence and reflects the administration’s broader effort to align federal funding instruments with national AI policy objectives.
Organizations pursuing federal education grants involving AI components should review the published priority and definitions to confirm alignment, particularly where grant proposals describe data collection, algorithmic decision support, or workforce training pipelines. State and local education agencies should anticipate that federal grant criteria will increasingly require demonstrated AI governance practices, including alignment with the NIST AI Risk Management Framework, and should establish documentation practices now rather than at the point of application.
Tags: Guidance | European Union
Date: May 12, 2026
Legal analyses of EDPB Guidelines 1/2026 highlight tighter expectations for AI-enabled research before June consultation close
Sidley Austin’s Data Matters privacy blog published a detailed legal analysis on May 12, 2026 of the European Data Protection Board’s Guidelines 1/2026 on the processing of personal data for scientific research purposes, the most significant EU regulatory statement on research data since GDPR took effect. The Guidelines, adopted by the EDPB on April 15, 2026, are open for public consultation until June 25, 2026, and are expected to be finalized in substantially their current form.
The Guidelines establish six indicative factors for determining whether an activity qualifies as scientific research under GDPR, including methodical approach, adherence to ethical standards, verifiability, autonomy, and contribution to general scientific knowledge. The EDPB confirms that broad consent and dynamic consent remain valid bases for research processing but only with documented safeguards under Article 89(1), including independent oversight, secure processing environments, privacy-enhancing technologies, and proactive participant communication. For AI-enabled research and large-scale processing of genetic or health data, a Data Protection Impact Assessment will almost always be required under Article 35.
Organizations conducting AI-enabled research using EU resident data should map current safeguards against the EDPB’s catalogue and conduct a gap analysis before the Guidelines are finalized. Multi-party research arrangements should be reviewed for controller and processor role allocation, particularly where sponsors access only pseudonymized data but determine processing purposes. Organizations with sector-specific concerns should submit consultation feedback before the June 25 deadline; the consultation window is the final structured opportunity to shape the finalized text.
AI INDUSTRY
Tags: News | Industry | Security
Date: May 12, 2026
Microsoft unveils MDASH agentic security system after finding 16 Windows vulnerabilities
Microsoft announced on May 12, 2026 that its new multi-model agentic scanning harness, codenamed MDASH, was used to discover 16 previously undisclosed vulnerabilities in the Windows networking and authentication stack, four of which are rated critical and were addressed in the May 2026 Patch Tuesday release. The discovered vulnerabilities span tcpip.sys, ikeext.dll, http.sys, dnsapi.dll, netlogon.dll, and telnet.exe; most are reachable from a network position without credentials.
MDASH was built by Microsoft’s Autonomous Code Security team and orchestrates more than 100 specialized AI agents across an ensemble of frontier and distilled models to discover, debate, and prove exploitable bugs end to end. Microsoft reported that MDASH achieved 96 percent recall on five years of confirmed Microsoft Security Response Center cases for clfs.sys and 100 percent recall on tcpip.sys, and scored 88.45 percent on the public CyberGym vulnerability discovery benchmark, placing it five points ahead of the next entry. The announcement follows Anthropic’s Project Glasswing and OpenAI’s Daybreak, both AI-powered cybersecurity initiatives targeting the same vulnerability discovery problem.
The strategic implication for enterprise security programs is that AI-driven vulnerability discovery has moved from research curiosity to production engineering, and the durable advantage lies in the agentic system around the model rather than any single model. Organizations should anticipate that the same capabilities accelerate adversarial discovery at comparable pace; defenders that have not implemented continuous patching pipelines aligned to vendor disclosure cycles will not absorb the increased disclosure tempo. MDASH is in limited private preview with select Microsoft customers ahead of broader rollout.
Tags: Alert | Industry | Security
Date: May 13, 2026
Eighteen-year-old NGINX rewrite module flaw enables unauthenticated remote code execution
Cybersecurity research firm depthfirst disclosed CVE-2026-42945 on May 13, 2026 in coordination with F5, a heap buffer overflow vulnerability in the ngx_http_rewrite_module of NGINX Plus and NGINX Open Source that has remained undetected for 18 years. The flaw, dubbed NGINX Rift, carries a CVSS v4 score of 9.2 and allows an unauthenticated attacker to crash worker processes or achieve remote code execution on systems with ASLR disabled. NGINX powers an estimated one-third of the world’s web servers, making the exposure surface significant.
The vulnerability is reachable when a rewrite directive combines unnamed PCRE captures with a replacement string containing a question mark, followed by a rewrite, if, or set directive in the same scope. The root cause is a state mismatch in NGINX’s two-pass script engine: the destination buffer is allocated under one set of escaping assumptions, then written under another, allowing attacker-controlled URI data to overflow the heap. Affected versions include NGINX Open Source 0.6.27 through 1.30.0 and NGINX Plus R32 through R36; fixes are available in NGINX Open Source 1.30.1 and 1.31.0, and in NGINX Plus R32 P6 and R36 P4.
Organizations operating NGINX in production should apply the patches immediately and audit web tier configurations for rewrite directives matching the vulnerable pattern. Where patching cannot be completed in a defined window, organizations should replace unnamed captures with named captures in every affected rewrite directive as a configuration-level mitigation. The disclosure should also prompt a broader review of long-lived infrastructure dependencies; foundational web and proxy components receive limited adversarial scrutiny relative to their deployment scale, and the assumption of safety derived from age is not a defensible control.
Source: https://thehackernews.com/2026/05/18-year-old-nginx-rewrite-module-flaw.html
