Nov 2: Three chained Ivanti zero-days give unauthenticated root access
Today's 3 things that matter
- Three chained Ivanti zero-days give unauthenticated root access
VPN gateway compromise enables lateral movement, credential theft (LAPS, BitLocker), and persistent network access; network teams must patch immediately and forensically investigate systems for backdoors and credential exfiltration. - eMudhra intermediate keys had as little as three bits entropy
Certificate authority compromise exposes encrypted TLS traffic to interception; infrastructure teams must audit certificate chains, update OS trust stores, and verify certificate validation across internal PKI and cloud services immediately. - Marvell to buy Lumentum in $115 billion all-stock deal
The deal signals that network infrastructure for AI data centers is shifting away from copper toward optical connectivity due to growing AI data traffic and physical limits of conventional wiring—directly affecting how you architect AI cluster networking and interconnect planning.
Full stories below, grouped by topic.
Network Automation
Itential outlines governance rules for agents acting on networks
Itential · Oct 26, 2026 · Primary source
What happened: Itential published guidance on agentic network operations, where AI agents execute network tasks under governance controls. The framework separates agents from chatbots: agents take an operational goal, plan steps, run them with audit trails, and iterate until the goal is met—but do not replace Ansible, Python, NSO, or monitoring tools.
Why it matters: Defines how practitioners should architect AI into network automation: agents orchestrate reasoning, but governed automation platforms execute deterministically. Clarifies what agents actually replace in your stack.
Autonomous network operations are narrow today but expanding. Gartner projects AI agents will be the most common way to execute network runtime activities by 2030. Itential's definition is precise: an agent is not a chatbot or single-purpose script. It takes an operational goal, plans the steps required, runs them under governance (with access controls, approvals, and audit trails), checks the result, and keeps iterating until the goal is met. Critically, agents do not replace Ansible, Python, NSO, monitoring systems, AIOps platforms, or network source-of-truth data—they sit on top of them. A network becomes autonomous one task at a time; the operating model is to measure how an automated task behaves, let results decide what runs autonomously next. This is material for practitioners building autonomous networks: it's about layering agentic reasoning on top of proven infrastructure automation, not rip-and-replace.
Read the original at itential.com
AIOps & Network Observability
Forrester names Dynatrace a leader in AIOps platforms
Dynatrace · Oct 30, 2026 · Vendor release
What happened: Forrester named Dynatrace a Leader in AIOps platforms, scoring highest in Current Offering category across 24 evaluation criteria. Dynatrace received the highest scores in 16 individual criteria, emphasizing causal AI for root cause analysis and unified observability.
Why it matters: Validates causal AI approach for deterministic root cause analysis in enterprise AIOps; shifts from correlation to causality for faster MTTR and trustworthy automation in NOC operations.
Forrester evaluated 10 AIOps vendors on 24 criteria including AI capabilities, observability features, and automation. Dynatrace's Davis AI engine uses causal analysis rather than correlation, distinguishing symptom chains from root causes across distributed systems. The platform combines real-time unified context with deterministic reasoning to surface actionable insights NOC teams can act on without guesswork. Dynatrace emphasizes that AI in IT operations is only useful when teams trust it; causal analysis grounds automation in explainable logic, critical for production incident response. The Leader designation reflects Dynatrace's focus on preventive operations and autonomous SRE capabilities, where automated decisions must be auditable. For network operations, this means incident correlation becomes mathematically traceable rather than heuristic-based, reducing false positive escalations and enabling safe automation at scale.
Read the original at dynatrace.com
Data Center & AI Networking
Oracle adds multiplanar RDMA fabrics to Acceleron networking suite
Oracle Cloud Infrastructure Blog · Oct 27, 2026 · Primary source
What happened: Oracle announced new capabilities in Oracle Acceleron, OCI's suite of networking and infrastructure technologies for demanding AI workloads. The expansion adds RDMA-based fabrics, resilient multiplanar networking and distributed collective communications for improved scalability and reliability.
Why it matters: OCI's Oracle Collective Communications Library (OCCL) plugin enables multi-planar fabrics without requiring specialized multi-port NICs, giving practitioners flexible topology designs and better NIC utilization for large distributed AI clusters.
Oracle Acceleron extends front-end RDMA to inference, storage, and HPC workloads, moving data directly between memory regions while slashing CPU overhead. The new OCCL plugin resolves a critical architectural constraint by enabling multi-planar networking for general AI/ML workloads without dedicated multi-port NIC requirements. Built-in failover, recovery, and congestion-aware load balancing keep multi-cluster AI training jobs running efficiently despite network congestion or failures. OCI's front-end RDMA supports all Queue Pair types and verbs without application changes, allowing customers to run storage, AI, and general-purpose traffic on the same frontend network while preserving performance isolation. This approach enables practitioners to connect separate GPU clusters over front-end lanes while backend fabrics handle local training, creating flexible multi-cluster architectures for complex distributed AI workloads.
Read the original at blogs.oracle.com
Routing & Internet
Cloudflare tracks exploitation of weak eMudhra certificate keys
Cloudflare Blog · Oct 30, 2026 · Primary source
What happened: Four weak keys from eMudhra certificates can compromise unpatched connections using eMudhra-issued certificates; Cloudflare details patching actions, exploitation attempts observed on its network, and remaining Internet exposure.
Why it matters: Network and security operators must audit and patch systems using eMudhra-issued certificates; weak CA keys enable MITM attacks on encrypted sessions passing through unpatched infrastructure.
Four eMudhra keys are weak enough to break connections; seven additional similarly weak keys were identified (56, 66, 115, 155, 156, 230, 237). All high-priority Cloudflare systems have been patched, with a small number of confirmed low-impact tickets still being worked on. About 160 handshakes observed on Cloudflare's network stood out where certificates' Subject Alternative Names named unrelated organizations' domains; forged certificates bore names of consumer IoT domains, were issued after keys leaked, and remained valid for over two years—longer than any public CA permit. Operators should audit eMudhra certificates in TLS termination points, CDN edge configs, and enterprise gateways; evaluate exploitation risk based on certificate scope and network position; and coordinate rapid patching of affected systems before attackers leverage persistent forged credentials.
Read the original at blog.cloudflare.com
Network Security
eMudhra intermediate keys had as little as three bits entropy
Tech Insider · Oct 29, 2026 · Industry news
What happened: Indian Certificate Authority eMudhra issued eight TLS intermediate certificates with private keys bearing only 3–8 bits of entropy instead of the required 384 bits, enabling trivial key recovery and adversary-in-the-middle attacks. eMudhra revoked all eight CAs on October 18 and announced voluntary exit from global SSL/TLS business on October 22.
Why it matters: Certificate authority compromise exposes encrypted TLS traffic to interception; infrastructure teams must audit certificate chains, update OS trust stores, and verify certificate validation across internal PKI and cloud services immediately.
On October 17, 2026, a security researcher identified cryptographic weakness in eight eMudhra intermediate certificates. Chrome and Firefox blocked the certificates within hours of notification on October 17 evening. Mozilla and Google's Chrome Root Program confirmed the compromise on October 20. Microsoft released emergency out-of-band updates KB5134133 and KB5134125 on October 28 to remove eMudhra root trust. The compromise escalated to global root certificate distrust across Mozilla, Chrome, Apple, and Microsoft. No TLS certificates chaining to the six affected website intermediates had appeared in Certificate Transparency logs before Chrome's block, preventing unauthorized certificate use via CT logs. Teams must immediately audit all eMudhra-issued certificate chains, update operating system trust stores, review active certificate validation mechanisms, and verify that internal PKI infrastructure and cloud services do not rely on eMudhra-signed certificates for critical encrypted channels.
Read the original at tech-insider.org
Three chained Ivanti zero-days give unauthenticated root access
The Hacker News · Oct 31, 2026 · Industry news
What happened: Threat actor UTA0576 exploits three chained Ivanti Connect Secure flaws to achieve unauthenticated root-level remote code execution: CVE-2026-104935 (authentication bypass + RCE), CVE-2026-111357 (privilege escalation, CVSS 8.8), and CVE-2026-118527 (file upload). Exploited in the wild since October 3, 2026.
Why it matters: VPN gateway compromise enables lateral movement, credential theft (LAPS, BitLocker), and persistent network access; network teams must patch immediately and forensically investigate systems for backdoors and credential exfiltration.
Volexity discovered active exploitation by UTA0576 starting October 5, 2026, and disclosed it October 13. CVE-2026-104935 was patched October 13, but CVE-2026-111357 privilege escalation remained unpatched until October 29—a 16-day window during which attackers could gain full root access via the exploit chain. CISA added CVE-2026-104935 to the Known Exploited Vulnerabilities catalog October 13. After exploitation, attackers targeted Active Directory credentials, LAPS passwords, and BitLocker recovery keys. Ivanti's Integrity Checker Tool cannot detect CVE-2026-104935 exploitation because commands execute in memory. Federal civilian agencies must patch CVE-2026-111357 by November 2 and CVE-2026-118527 by November 13. Affected systems should be assumed compromised before October 13 and require forensic investigation for lateral movement and credential compromise.
Read the original at thehackernews.com
Startups & Funding
Marvell to buy Lumentum in $115 billion all-stock deal
TechXplore · Nov 2, 2026 · Industry news
What happened: Marvell Technology will acquire optical components maker Lumentum in a $115 billion all-stock deal, integrating networking chips and custom processors with optical transceivers and switching technology for AI data center infrastructure.
Why it matters: The deal signals that network infrastructure for AI data centers is shifting away from copper toward optical connectivity due to growing AI data traffic and physical limits of conventional wiring—directly affecting how you architect AI cluster networking and interconnect planning.
Marvell's $115 billion acquisition of Lumentum integrates two critical AI infrastructure layers: Marvell's networking chips and custom processors with Lumentum's lasers, optical transceivers and optical switching technology. Nvidia invested $2 billion in each company earlier in 2026 as part of a broader push to distribute its profits across the AI infrastructure stack. This consolidation accelerates the industry's move toward optical networking in hyperscale data centers. For practitioners building AI infrastructure, this signals that optical interconnect design choices—long considered advanced—are now becoming baseline requirements for scale-out and scale-up clusters. The deal reflects growing consensus that bandwidth density and latency constraints in conventional Ethernet and copper wiring constrain AI model training throughput at scale. Marvell's leadership frames this as responding to a fundamental architectural shift in AI infrastructure toward optical connectivity, with growing AI data traffic and physical limits of wiring driving the transition.
Read the original at techxplore.com
Serval raises $150 million for IT automation agents
SiliconANGLE · Oct 29, 2026 · Industry news
What happened: Enterprise service management startup Serval raised $150 million at a $3 billion valuation to expand its automation agents beyond IT operations, including new Catalyst agent for scanning support tickets and auto-generating workflows.
Why it matters: Serval's code generation tools for non-engineering staff and scaled on-site automation engineering teams are relevant if you're evaluating enterprise automation platforms for network operations transformation beyond traditional network engineering.
Serval announced a $150M Series C at a $3B valuation led by Sequoia Capital, bringing total funding to $277M in three rounds over 12 months. The funding supports Catalyst, an agent that scans support tickets for repetitive workflows and auto-generates code that administrators review before deployment, positioning automation as a tool for non-engineers. Serval is launching ProServal, a professional services unit hiring 100+ automation engineers to deploy on-site at large customers for multi-team workflow mapping and rapid automation builds, with about half being recent graduates in training. While Serval is primarily an enterprise service management platform rather than pure network automation, the automation-agent architecture and non-engineer tooling represent patterns increasingly relevant to NetOps teams building self-service network change automation.
Read the original at siliconangle.com
Agentic AI & MCP
Agent-to-agent protocols converge as MCP ships broadly
Mean CEO · Nov 2, 2026 · Analysis
What happened: November 2026: agent definitions stabilized across Google Cloud, IBM, Microsoft, GitHub, ServiceNow, BCG; MCP protocol shipping; agent-to-agent communication emerging as standard. 65% of FTSE 100 firms hiring for agent roles; 80% report skills gap.
Why it matters: Agent definitions and interoperability standards are crystallizing—NetOps platforms can now assume MCP compatibility and expect agent-to-agent orchestration to become standard architectural pattern.
By November 2026, the agent market has moved past exploration into standardization. Google Cloud, IBM, BCG, Microsoft, GitHub, ServiceNow and Informatica now describe autonomous AI agents in nearly identical terms: systems that execute toward goals autonomously, maintain persistent context, call external tools via standard protocols, and coordinate with other agents. This convergence signals maturity. MCP is shipping across platforms; A2A (agent-to-agent) communication is emerging—ServiceNow's platform now highlights agents talking to other agents natively. GitHub documents security automation agents detecting and fixing vulnerabilities across the software lifecycle. Agent orchestration is becoming infrastructure-level, not application-level. The talent signal: 65% of FTSE 100 companies have open positions for AI agent roles, but 80% report a skills gap finding candidates with both business acumen and technical depth to design intent, govern execution, and handle exceptions.
Read the original at blog.mean.ceo
Telco & Cable AI
Singtel runs customer calls on sovereign on-premise inferencing platform
Computer Weekly · Oct 26, 2026 · Industry news
What happened: Singtel built an on-premise sovereign inferencing platform consolidating years of scattered AI work, enabling its virtual agent to handle most customer calls without human help. The platform lets Singtel avoid cloud dependencies and retain vendor flexibility while scaling AI quickly.
Why it matters: On-premise, governed AI platforms let carriers move beyond single-vendor lock-in while meeting data residency requirements—a practical model for network operators managing critical infrastructure.
Singtel built an on-premise sovereign inferencing platform consolidating years of scattered AI work, enabling the telco to scale AI quickly without cloud lock-in. As a critical infrastructure provider, Singtel cannot load all AI use cases on public cloud. The platform lets the telco plug and play technology as it evolves rather than getting locked in with one vendor. Singtel has identified 10 areas where AI creates measurable value, rolling out AI agents across mobile, TV and fiber networks to assist engineers with root cause analyses in partnership with Ericsson and Nokia. The AI agents do not operate autonomously due to the risk of making wrong changes to the network. This approach balances automation gains with operational safety—a critical lesson for network practitioners deploying agentic systems into production infrastructure.
Read the original at computerweekly.com
TRAI drafts AI governance rules for Indian telecom operators
TelecomLead · Nov 1, 2026 · Industry news
What happened: TRAI released draft AI guidelines on October 30, 2026, as Indian telecom operators accelerate AI adoption across 5G network management, customer service, fraud prevention, network optimization, predictive maintenance and automated business operations. The proposed guidelines ensure that telecom companies deploy AI technologies responsibly while maintaining network reliability, consumer protection, data security and operational resilience.
Why it matters: Regulatory frameworks for telecom AI deployment are now mandating governance, auditability and resilience guardrails—network practitioners must expect similar compliance requirements on AI-native systems to become standard across jurisdictions.
TRAI's draft framework, released on October 30, 2026, comes as Indian telecom operators accelerate AI adoption across 5G network management, customer service, fraud prevention, network optimization, predictive maintenance and automated business operations. The proposed guidelines aim to ensure that telecom companies deploy AI technologies responsibly while maintaining network reliability, consumer protection, data security and operational resilience. TRAI has invited industry stakeholders to submit comments by November 27, 2026. Telecom companies increasingly depend on AI-powered platforms to automate network operations, predict equipment failures, optimize spectrum utilization and improve customer experiences. Telecom operators can use machine learning and predictive analytics to identify congestion, allocate network resources, forecast traffic demand and improve service availability. This regulatory push signals that governance frameworks for agentic AI in telecom are moving from optional best practice to compliance requirement—particularly critical for operators deploying autonomous decision-making at scale.
Read the original at telecomlead.com
AI Industry & Policy
Anthropic reports $16.3 billion first-half revenue in IPO filing
AI News for October 30, 2026 (SDG Strategy / Faithful Intelligence) · Oct 30, 2026 · Primary source
What happened: Anthropic filed to go public on October 26 with revenue hitting $16.3B in H1 2026, up from $1.2B a year earlier, with 11,900+ customers paying $100,000 or more annually. Nearly half its revenue flows through Amazon, Google and Microsoft's cloud arms.
Why it matters: Enterprise adoption is accelerating with major revenue scale and customer concentration at the $100K+ tier, establishing baseline consumption patterns for production AI workloads.
Anthropic filed to go public on Nasdaq as 'ANT' on October 26. Revenue rose dramatically to $16.3B in H1 2026 from $1.2B in the prior-year period, indicating 13.6x growth. Over 11,900 customers pay $100,000 or more annually. Nearly half its revenue flows through Amazon, Google and Microsoft's cloud arms, revealing enterprise distribution concentration and the outsized leverage cloud providers hold over model access. This represents the first disclosed revenue scale for a frontier AI lab and establishes benchmarks for what production enterprise adoption looks like at scale. The customer base composition—concentrated in large-value contracts rather than breadth—signals that AI deployment remains capital-intensive and limited to well-funded enterprises.
Read the original at aisdg.substack.com
Canada takes ownership stake in AI company Cohere
BNN Bloomberg Canada · Nov 1, 2026 · Industry news
What happened: Canada acquired an ownership stake in Cohere, a homegrown AI company competing with American and Chinese AI giants. In September, Cohere merged with Berlin-based Aleph Alpha to create the first transatlantic sovereign AI solution.
Why it matters: National governments are now directly investing in domestic AI champions as strategic assets to reduce dependency on U.S. and Chinese technology and prevent geopolitical leverage through AI access.
Canada is taking a step toward digital sovereignty after acquiring an ownership stake in Cohere, a homegrown AI company that wants to take on American and Chinese AI giants. In September, Cohere merged with Berlin-based Aleph Alpha to create the first transatlantic sovereign AI solution, combining Canadian and European resources. Major investors include Germany's Schwarz Group, NVIDIA, and Bell Ventures, with Cohere saying government funding will allow it to create 1,000 new Canadian technology jobs. Analysts say supporting Cohere provides an alternative to using American and Chinese AI and reduces the risk of those countries using the technology as leverage. This consolidation signals a shift toward state-backed industrial policy for AI as core infrastructure, mirroring patterns in the U.S., EU, and Asia. The Cohere-Aleph Alpha merger specifically creates a transatlantic alternative positioned against OpenAI and Anthropic.
Read the original at bnnbloomberg.ca
Read this edition on the web · The week in network intelligence · Vendor Radar
Digital Plumber is AI-curated and AI-summarized, with no human review before publishing. Verify before acting on anything here. How it works.