You're reading the Friday, November 13, 2026 edition. Today's briefing →
Toronto
Friday, November 13, 2026No. 152

Digital Plumber

Plumbing the information age

AI-curated intelligence for people who run networks. Daily coverage of AIOps, network automation, agentic operations, AI infrastructure, security and the vendors shaping them.

Today's 3 things that matter

Picked by the AI editor
  1. Security·Industry news

    Critical FortiOS routing flaw and FortiManager RCE patched

    Fortinet released November 2026 advisories fixing a critical FortiOS routing flaw (CVE-2026-126253, CVSS 9.3) enabling unauthenticated attackers to trigger firewall routing announcements, and critical FortiManager flaws including unauthenticated code execution (CVE-2026-126261, CVSS 9.0) via the FGFM management channel.

    Why it matters Network operators running FortiOS or centralized FortiManager must assess attack surface; routing flaws require OSPF network proximity but JSON-RPC API flaws expose systems if API is enabled.

  2. Routing·Industry news

    BGP hijack via AS 402425 redirects Chinese traffic to gambling sites

    Malicious actors exploited BGP longest-prefix-match rules to announce false routes through AS 402425 (registered to US-based Qorexa Labs in April 2026), hijacking traffic across Chinese provinces and substituting DNS servers without infecting end-user devices.

    Why it matters Demonstrates continued BGP hijack risk despite RPKI; highlights operator filtering gaps and AS registration abuse as a vector for regional traffic interception.

  3. Agentic AI·Vendor release

    Over half of organizations run agentic AI in production NetOps

    Over half of organizations already use agentic AI to act on NetOps processes and tasks in production, per a 2026 Omdia/Cisco study of 1,000 IT leaders.

    Why it matters As network teams hand authority to AI agents, the challenge is deciding what to trust them with and ensuring verification costs less than manual work—a practical governance model for production NetOps.

Today's briefing

What happened, and why it matters

12 stories · 7 topics · Updated 12:45 PM ET

ThousandEyes traces Microsoft 365 Exchange Online outage across global network

ThousandEyes Blog · Nov 12, 2026 · Primary source

ThousandEyes published real-time outage analysis of the Microsoft 365 Exchange Online incident, tracking timeouts, packet loss, and traffic patterns across the global network starting at 17:30 UTC on November 12. The analysis demonstrates internet-scale observability and cross-layer correlation of network and application failures.

Why it matters Shows how modern NOC teams use internet-scale telemetry to detect and validate outages independent of vendor status pages, enabling faster incident response and root cause validation.

ThousandEyes' published analysis documented the Microsoft 365 outage starting at 17:30 UTC on November 12, 2026, when its global monitoring detected TCP connection resets, timeouts, and internal packet loss across Microsoft's infrastructure. The telemetry showed Exchange Online traffic experiencing significant degradation, with continued impact through the recovery period on November 13. ThousandEyes' Internet Insights service, which aggregates anonymized data across its customer base to detect broader internet disruptions, provided independent validation of outage scope and timeline—confirming the incident affected services globally well before Microsoft's formal status updates. The analysis illustrates the operational value of deterministic network observability: NOC teams using ThousandEyes could correlate application-layer failures (Outlook timeouts, 503 errors) with network-layer symptoms (packet loss, connection resets), providing evidence-based incident validation without relying on vendor statements. This real-world outage demonstrates how observability tools enable faster escalation decisions and verify when problems are resolved.

Read the original at thousandeyes.com ↗

Exchange Online traffic fell 85 percent during Microsoft outage

Kentik Internet Analysis · Nov 13, 2026 · Primary source

Kentik's Director of Internet Analysis published real-time traffic analytics showing Exchange Online traffic fell 85% below expected levels at 17:35 UTC on November 12, providing independent quantification of outage severity. Recovery began at 11:20 UTC on November 13, with queue catch-up traffic running 19% above baseline by 13:25 UTC.

Why it matters Demonstrates how network observability platforms provide quantified, time-stamped evidence of service impact severity—critical for SLO validation, incident severity classification, and post-incident root cause investigations in multi-tenant environments.

Kentik's analysis, led by Doug Madory (Director of Internet Analysis), tracked the Microsoft Exchange Online outage using BGP-level and application-layer network telemetry. The analysis quantified the incident's scope: Exchange Online traffic dropped from expected levels to 85% below baseline starting at approximately 17:35 UTC on November 12, providing objective proof of outage onset independent of user reports or internal Microsoft metrics. Kentik's continuous monitoring captured the recovery phase on November 13, showing traffic returning to normal at 11:20 UTC, with subsequent queue recovery pushing traffic 19% above expected as delayed messages caught up. For NOC teams managing SLAs, this independent quantification enables non-repudiation: the traffic data serves as forensic evidence of when services actually failed and recovered, separate from any internal or vendor claims. The ability to measure impact in network-native metrics (traffic volume, BGP routing changes, latency) gives operations teams a technical foundation for escalation decisions, customer communications, and post-incident reviews—especially critical when vendors like Microsoft face scrutiny over timeline accuracy during major incidents.

Read the original at ppc.land ↗

Most firms run AI agents in operations without execution transparency

Dynatrace · Nov 12, 2026 · Research

Dynatrace research in its State of Observability 2026 report finds that 87 percent of technology leaders use AI agents for observability and IT operations, yet controls needed to verify what those agents do haven't kept pace. The study identifies a 14-percentage-point gap between AI agent deployment and execution transparency.

Why it matters Highlights a critical challenge for AIOps practitioners: while AI agents are deployed widely, organizations lack observability and governance mechanisms to validate agent decisions—a prerequisite for autonomous remediation at scale in critical infrastructure.

Dynatrace research published November 12, 2026, found that 87 percent of technology leaders already deploy AI agents for observability and IT operations, yet the controls needed to verify and govern agent actions haven't kept pace with deployment. This gap manifests across nine measured AI observability capabilities, with execution trails perceived importance lagging actual deployment by an average of 14 percentage points—spanning quality assurance, agent decision visibility, security, cost attribution, and business impact assessment. For NOC and NetOps teams, this signals a maturity challenge: autonomous automation deployment has outpaced ability to audit, validate, and govern autonomous actions in production. Integration complexity, skills shortages, data quality concerns, and explainability requirements all emerge as barriers to greater autonomy. Dynatrace positions its Smartscape unified topology view and Davis AI causal reasoning engine as solutions, enabling topology-based root cause analysis that traverses infrastructure relationships, communications, and dependencies to establish deterministic causality rather than probabilistic correlation—addressing the need for explainable, auditable autonomous decision-making in critical infrastructure operations.

Read the original at developer-tech.com ↗

BGP hijack via AS 402425 redirects Chinese traffic to gambling sites

tribuna.com · Nov 11, 2026 · Industry news

Malicious actors exploited BGP longest-prefix-match rules to announce false routes through AS 402425 (registered to US-based Qorexa Labs in April 2026), hijacking traffic across Chinese provinces and substituting DNS servers without infecting end-user devices.

Why it matters Demonstrates continued BGP hijack risk despite RPKI; highlights operator filtering gaps and AS registration abuse as a vector for regional traffic interception.

Malicious actors exploited BGP rules to announce false routes through foreign AS 402425, registered in April 2026 by Qorexa Labs. The attack leveraged longest-prefix match to override legitimate routes, substituting official website addresses with gambling-site destinations or malicious DNS servers without infecting end-user devices. While measurements pinpointed tampering within AS 402425, investigators noted that analysis does not establish intent on the part of the company. CNCERT urged international registries and IP-address trading platforms to review and revoke malicious AS resources, while recommending that operators filter out routes containing AS 402425. This incident reinforces the need for strict egress filtering at regional boundaries and rapid coordinated response to malicious ASNs.

Read the original at tribuna.com ↗

eMudhra revokes eight more CAs after private key blunder

BleepingComputer · Nov 11, 2026 · Industry news

eMudhra revoked eight additional intermediate CA certificates after discovering improper handling of private keys during certificate generation, where a script reduced eight ECDSA private keys to a single byte. Chrome and Firefox will stop trusting emSign G1-chained certificates as of November 16, 2026.

Why it matters Certificate holders must immediately reissue TLS certificates from unaffected CAs to avoid trust failures; infrastructure teams need immediate CA trust store and certificate inventory audits.

Following October's discovery of 8 eMudhra intermediate CAs with trivially weak keys, the November investigation found a second round of CA compromise: 8 RSA intermediate CAs exposed to process memory during HSM transfers due to a flawed in-house Python script. While these 8 RSA keys were not cryptographically weakened like the October batch, they were nonetheless revoked November 7 on advice from browser root programs. The dual compromise now totals 16 affected intermediate CA certificates. Chrome, Firefox, and Safari are removing emSign G1 root certificates from their trust stores on a phased basis: Chrome and Firefox November 16, Apple November 30. Organizations holding certificates chaining to emSign G1, emSign ECC G3, or G2C intermediates face either immediate replacement or certificate rejection by browsers and TLS clients. This represents a systemic PKI failure affecting tens of thousands of organizations globally, particularly in higher education and government sectors. Infrastructure teams must prioritize certificate reissuance from DigiCert, Sectigo, or other established CAs, and verify certificate chains in production systems before November 16 cutoff dates to prevent service outages.

Read the original at bleepingcomputer.com ↗

Critical FortiOS routing flaw and FortiManager RCE patched

Red Tiger Security · Nov 10, 2026 · Industry news

Fortinet released November 2026 advisories fixing a critical FortiOS routing flaw (CVE-2026-126253, CVSS 9.3) enabling unauthenticated attackers to trigger firewall routing announcements, and critical FortiManager flaws including unauthenticated code execution (CVE-2026-126261, CVSS 9.0) via the FGFM management channel.

Why it matters Network operators running FortiOS or centralized FortiManager must assess attack surface; routing flaws require OSPF network proximity but JSON-RPC API flaws expose systems if API is enabled.

Fortinet's November 10 release addresses 17 CVEs with three critical for network infrastructure. CVE-2026-126253 (OSPF auth bypass, CVSS 9.3) allows unauthenticated attackers on the same network segment running OSPF to craft packets causing unauthorized routing announcements—a control-plane threat for internal network hijacking. CVE-2026-126170 (JSON-RPC JSON injection, CVSS 9.8) affects FortiManager and FortiAnalyzer when JSON API is enabled, allowing unauthenticated code execution on the management console. CVE-2026-126261 (FGFM service auth bypass, CVSS 9.0) enables attackers reaching the FortiGate-to-FortiManager port to execute code. As of November 12, none show confirmed exploitation in the wild and none appear in CISA KEV, but critical severity makes these priority. Network teams should inventory FortiManager deployments and JSON API enablement status before scheduling patches. Routing flaws affect only MX Series hardware and specific HA configurations, reducing blast radius compared to centralized management vulnerabilities.

Read the original at redtigersecurity.com ↗

CISA adds exploited OpenSSH and FreeSwitch flaws to KEV

The Hacker News · Nov 9, 2026 · Industry news

CISA added CVE-2026-123158 (OpenBSD OpenSSH SSHole authentication bypass, CVSS 8.1) and CVE-2026-116779 (SignalWire FreeSwitch code injection, CVSS 9.8) to the KEV catalog on November 9. Exploitation of SSHole was first observed in honeypots November 7, 2026, with active attacks ongoing since November 6.

Why it matters Infrastructure teams must immediately patch OpenBSD OpenSSH on Linux systems; exploitation is currently active with honeypot hits confirmed by GreyNoise. Any internet-exposed SSH services need emergency patching.

CVE-2026-123158 (SSHole, CVSS 8.1) is an unexpected status code or return value vulnerability in OpenBSD OpenSSH on Linux enabling authentication bypass. GreyNoise documented first exploitation on November 7 at 12:46 UTC with activity accelerating since November 6. The flaw affects OpenBSD OpenSSH packaged for Linux distributions; internet-facing SSH services need immediate patching with November 2026 fixes available. CVE-2026-116779 (SignalWire FreeSwitch, CVSS 9.8) is a code injection vulnerability enabling configuration disclosure, credential theft, arbitrary call control, and remote command execution—critical for VoIP environments. CVE-2026-102895 (GPUStack, CVSS 9.8) and CVE-2026-123218 (IBM Aspera Faspex, CVSS 9.8) represent broader authentication and authorization bypasses affecting ML infrastructure and enterprise file transfer. Network teams should prioritize OpenSSH patching as immediate action 1, then FreeSwitch if VoIP is deployed, followed by other KEV entries on risk-based schedule aligned with system criticality and internet exposure.

Read the original at thehackernews.com ↗

Over half of organizations run agentic AI in production NetOps

Harvard Business Review / Cisco · Nov 11, 2026 · Vendor release

Over half of organizations already use agentic AI to act on NetOps processes and tasks in production, per a 2026 Omdia/Cisco study of 1,000 IT leaders. Enterprise monitoring systems generate over 4,000 alerts daily, roughly half network-related, driving the shift to autonomous triage and action.

Why it matters As network teams hand authority to AI agents, the challenge is deciding what to trust them with and ensuring verification costs less than manual work—a practical governance model for production NetOps.

The Cisco/Omdia research reveals rapid adoption of agentic NetOps in production environments. Half of surveyed organizations are already deploying AI agents to handle network operations tasks autonomously. The core business driver is alert fatigue: with 4,000+ daily alerts and growing complexity from AI workloads adding new traffic patterns and dependencies, manual triage is unsustainable. However, the research emphasizes a critical operational constraint: the cost of verifying an agent's work must be lower than performing the work manually. This frames the real decision boundary for NetOps teams—not whether to automate, but which tasks pass the verification cost test. The study shows that organizations are moving beyond pilot phases and deploying agents in production, suggesting frameworks for safe autonomy are emerging. The tension between speed and verification remains central to adoption rates.

Read the original at hbr.org ↗

Telcos move fastest on letting AI run customer journeys

Telecom's Tech News · Nov 12, 2026 · Analysis

Telecom companies are deploying AI to autonomously orchestrate customer journeys at a pace outpacing other industries. Over 90% of telco leaders expect to let AI independently select plans and renew contracts within two to three years, while 89% say AI assistants are replacing search engines as the starting point for customer journeys.

Why it matters Operations teams need to prepare BSS/OSS platforms for autonomous AI decision-making in customer lifecycle management, requiring new auditability and consent mechanisms for production deployments.

Telecom companies are deploying AI to autonomously orchestrate customer journeys at a pace that outstrips virtually every other industry, according to Acxiom's 2027 CX trends report. More than nine in ten telco leaders expect to be comfortable letting AI independently select plans and renew contracts within two to three years, a level of autonomy consumers haven't endorsed at comparable rates. Some 89 percent of telcos say AI assistants are replacing search engines as the starting point for customer journeys, and operators that don't actively work to influence those assistants risk losing visibility before a customer ever reaches a branded touchpoint. Regional expectations on timing diverge: most US telcos expect customers to accept AI optimizing services, identifying savings opportunities, and managing plan changes within five years. This gap between operator and customer trust in autonomous AI decision-making will require new approaches to transparency, auditability, and customer consent mechanisms within operational and customer-facing systems.

Read the original at telecomstechnews.com ↗

Google DeepMind ships Gemini Robotics 2.5 to robotics partners

CASRAI · Nov 13, 2026 · Industry news

Google DeepMind announced Gemini Robotics 2.5 on November 10, 2026, its most capable vision-language-action model that converts visual and language input into robot motor control. Rolling out to partners including Apptronik, Mercedes-Benz and Agile.

Why it matters Robotics operators now have access to a production-ready embodied AI model that translates vision and language into physical control, replacing custom end-to-end training for robotic tasks.

Google DeepMind announced Gemini Robotics 2.5 on November 10, 2026, describing it as its most capable vision-language-action (VLA) model to date. The model turns visual and language input into robot motor control. DeepMind's announcement says Gemini Robotics 2.5 is rolling out to robotics partners including Apptronik, Mercedes-Benz and Agile, though the announcement does not describe a public API. This represents a narrower, partner-focused deployment strategy compared to general-purpose language models, targeting specific robotics manufacturers and integrators rather than providing broad public developer access. The model represents advancement in enabling physical systems to understand and execute complex tasks from high-level descriptions.

Read the original at casrai.org ↗

Chip export politics reshape AI infrastructure funding and valuations

Foreign Affairs Forum · Nov 12, 2026 · Analysis

H200 chip licensing to China became a 2026 flashpoint with bipartisan congressional concern, while Anthropic raised $65 billion at a $965 billion valuation and Fractile, a London chip-design firm, attracted multiples-higher valuations after signing with Anthropic.

Why it matters Large American technology companies are widening capital sources across Asia during heavy computing investment, creating financial ties that interact with chip policy geopolitics—directly affecting enterprise infrastructure procurement and supply-chain risk.

This analysis examines how U.S.-China AI chip competition is reshaping capital flows and infrastructure strategy. Key dynamics: the H200 licensing debate between Commerce Department and Congress over Chinese AI access; Anthropic's $65 billion raise (among the largest of 2026) at a $965 billion valuation; and the emergence of alternative chip suppliers like Fractile (London), valued at roughly $1 billion in May 2026 but commanding multiples higher valuations by mid-year after securing supply deals with frontier AI labs. The piece highlights that geopolitical competition over AI architecture—not just raw chip availability—is driving investment patterns and reshaping technology partnerships. For infrastructure practitioners, this means chip sourcing, data center siting, and cloud procurement decisions now carry explicit geopolitical risk. The financing ties between U.S. tech companies and Asian investors add another layer: capital flows themselves become vectors of strategic competition.

Read the original at faf.ae ↗

Southeast Asia AI funding more than doubles to $4.2 billion

TNGlobal (Tracxn) · Nov 12, 2026 · Research

Southeast Asia AI companies raised $4.2 billion in the first nine months of 2026—up 121 percent year-over-year—with the largest round being Kling AI's $2.8 billion Series D. The ten largest rounds accounted for 98 percent of all SEA AI funding.

Why it matters SEA AI funding concentration is extreme: mega-rounds dominate (top 10 = 98% of capital), seed funding declined, and 100% of SEA AI funding came from Singapore-based companies—meaning infrastructure and adoption patterns are highly consolidated and geographically concentrated.

Tracxn's comprehensive 2026 data reveals structural shifts in Southeast Asia's AI ecosystem. The headline 121% year-over-year growth masks a capital concentration problem: Kling AI's $2.8B Series D (video/content), Sharpa's $670M Series D (backed by Alibaba, Meituan, Tencent), and Zelostech's $300M Series C dominated, while the 98% concentration in top-10 rounds signals that mid-stage and early-stage companies are starved of funding. Seed funding rose 28% in value but fell in count (10 rounds vs. 18 prior year), suggesting fewer early-stage opportunities. On the exit side: limited activity—one acquisition (Meta bought Manus in April 2026 for undisclosed sum), one IPO, no new unicorns. Investor count dropped from 69 to 56 year-over-year; 39 were first-time investors. Geographic concentration is near-total: 100% (vs. 99% prior year) came from Singapore-based companies, indicating negligible activity outside the city-state. For practitioners evaluating SEA as an AI adoption region, this concentration suggests infrastructure, talent, and vendor ecosystems are centered in Singapore with limited regional diversification.

Read the original at technode.global ↗
Nothing in today's briefing matches that.

Listening

Podcasts and talks
  • Packet Pushers

    Unimus To Add Native MCP Support – AutoCon 6 Briefing

    Unimus, a network configuration management platform, is integrating native Model Context Protocol (MCP) support, enabling AI agents direct access to network backup data and device configuration information through natural language queries.

  • Network to Code

    Building AI Factory Networks with Nautobot and NVIDIA Config Manager

    NVIDIA is automating AI infrastructure using Config Manager (NVCM), an open-source tool combining source-of-truth management, configuration rendering, ZTP, DHCP, and workflow automation—principles that apply equally to branch networks and data center fabrics.

  • Packet Pushers (Heavy Networking)

    HN850: AWS Takes a Unique Approach to AI Fabrics (Sponsored)

    AWS Distinguished Engineer and Network Core VP JR discusses the company's network approach to AI cluster connectivity, covering traffic patterns, topology design, and operational principles specific to AI workload networking.

  • Network to Code

    Last Month in Nautobot - October 2026

    Network to Code's monthly roundup covers Nautobot feature releases, community contributions, and product direction updates from October 2026, tracking incremental improvements to the network automation platform.

Vendor Radar

Last 7 days · arrows compare with the 7 before

What changed this week

Last 7 days vs the 7 before

Biggest moves

Trending topics

agent · observability · Agentic AI · automation · OpenTelemetry · RAG · AIOps · Kubernetes · inference · MCP · LLM · SRE

Get the daily briefing

Today's 3 things that matter and every story with why it matters, in your inbox each morning. Free, and you can unsubscribe at any time. Prefer a reader? Follow the RSS feed.