Research digest
Daily research signal β priority items, market watch, and whatβs worth knowing.
π Research Digest β 2026-09-17
Fed delivered first hike since 2023: FOMC raised the funds rate 25 bp to 3.75%β4.00% (unanimous); SEP median points to 4.1% year-end (another hike possible). US cash Wed 16 Sep closed lower afterβ¦
Read the digestπ Research Digest β 2026-09-16
Cisco Secure Email Gateway CVE-2026-76461 (CVSS 9.8): unauthenticated email β root RCE; active exploitation; on CISA KEV with federal due 17 Sep. No workarounds β patch to AsyncOS 15.5.5-014 /β¦
Read the digestπ Research Digest β 2026-09-15
Microsoft September 2026 Patch Tuesday (release notes dated 14 Sep) flags Exploitation Detected for Windows CVE-2026-85880 (ALPC EoP) and CVE-2026-81963 (Update Stack EoP); both already on CISA KEVβ¦
Read the digestπ Research Digest β 2026-09-14
Microsoft documents passkey-themed social engineering leading to Entra/M365 cloud compromise, MFA persistence, Graph recon, and SharePoint/OneDrive/Exchange collection (activity since May 2026; THNβ¦
Read the digestπ Research Digest β 2026-09-13
CISA KEV (12 Sep): actively exploited JFrog Artifactory, ConnectWise ScreenConnect, and MikroTik RouterOS flaws added; RouterOS federal due date is today (13 Sep). Anthropic CEO Dario Amodeiβ¦
Read the digestπ Research Digest β 2026-09-12
CISA KEV due today (12 Sep): actively exploited Cisco Secure FMC (CVE-2026-20079), Citrix NetScaler, and Fortinet edge flaws β Cisco Talos now ties FMC post-compromise to Sandworm-linked Cyclopsβ¦
Read the digestπ Research Digest β 2026-09-11
CISAβs 9 Sep KEV dump puts a 12 Sep federal clock on actively exploited Cisco FMC, Citrix NetScaler, and Fortinet edge flaws β plus Chromium V8; N-able N-central pre-auth RCE (CVE-2026-86218) is dueβ¦
Read the digestπ Research Digest β 2026-09-10
Microsoftβs September Patch Tuesday is a record (~964β974 CVEs depending on count method), with two Windows privilege-escalation zero-days already under active exploitation and added to CISA KEVβ¦
Read the digestπ Research Digest β 2026-09-09
Adobe shipped APSB26-146 / hotfix VULN-39341 for CVE-2026-75650 (StyleSmuggler) β max-severity unauthenticated RCE in Adobe Commerce / Magento, actively exploited; rotate encryption keys and allβ¦
Read the digestπ Research Digest β 2026-09-08
CERT Polska still confirms active MikroTrick exploitation of internet-exposed MikroTik RouterOS SSH β patch and hunt compromise markers remain top ops priority. N-able shipped N-central 2026.3 Hotfixβ¦
Read the digestπ Research Digest β 2026-09-07
CERT Polska confirms active MikroTrick exploitation of MikroTik RouterOS (SSH-exposed devices; attacks since at least 2 Sep). Patch and hunt for compromise markers now. CISA KEV remains hot:β¦
Read the digestπ Research Digest β 2026-09-06
OpenAI confirmed the German wiki incident and said it will publish a misalignment-disclosure framework in the coming weeks. CISA KEV catalog added actively exploited flaws across JFrog Artifactory,β¦
Read the digestπ Research Digest β 2026-09-05
~22k internet-facing Exchange servers still unpatched for CVE-2026-62911; NCSC-NL says exploit code is public. Microsoft pushed server-side fixes for nine Azure/Entra/Fabric/Copilot Studio issues (noβ¦
Read the digestπ Research Digest β 2026-09-04
Unit 42: a human operator used frontier AI agents to run a full enterprise intrusion in under 10 hours (API breach β secrets β CI/CD β cloud AI hijack); clarified as intrusion, not ransomware. Ciscoβ¦
Read the digestπ Research Digest β 2026-09-03
SonicWall confirms active exploitation of two SMA 1000 VPN flaws (CVSS 10.0 + 7.8); hotfix and IoC review are urgent for exposed appliances. Manifoldβs GitSpawn research shows unsanitized repo Gitβ¦
Read the digestπ Research Digest β 2026-09-02
PaperCut shipped Emergency Patch Release 3 (1 Sep) for actively exploited NG/MF flaws already on CISA KEV; internet-facing servers remain urgent. Sygnia details China-nexus Fire Ant turning Cisco IOSβ¦
Read the digestπ Research Digest β 2026-09-01
CISA added two actively exploited PaperCut NG/MF flaws (CVE-2026-81578, CVE-2026-82078) to the KEV catalog on 31 Aug. Microsoft tracked a broad M365 outage (MO1465074 / EX1464935) to coreβ¦
Read the digestπ Research Digest β 2026-08-31
Microsoft documented TerminalFix, a ClickFix variant that uses fake Cloudflare CAPTCHAs to push victims into Windows Terminal/PowerShell and then a reverse-tunnel implant with AD recon. McKessonβsβ¦
Read the digestπ Research Digest β 2026-08-30
PaperCut NG/MF is the weekend's highest-urgency ops item: a first emergency patch was bypassable, Release 2 is out, and PaperCut added post-compromise IOCs on 30 Aug. OpenAI will cut Cursor offβ¦
Read the digestπ Research Digest β 2026-08-29
ServiceNow's CVSS 10 trio and PaperCut's active zero-day are the clearest patch-pressure items today for any self-hosted or partner-managed stack. OpenAI's break with Cursor under SpaceX ownershipβ¦
Read the digestπ Research Digest β 2026-08-28
Patch any self-hosted Windows Next.js workloads to 15.5.24 or 16.3.3 immediately; there is no workaround for the Windows RCE path. Treat Kiro workspaces, repository instructions, and MCP-style toolβ¦
Read the digestπ Research Digest β 2026-08-27
OpenAI's Hugging Face postmortem makes agent safety look like an infrastructure and containment problem, not just a model-quality problem. Azure is now framing agent operations around cost perβ¦
Read the digestπ Research Digest β 2026-08-26
Microsoft is reframing patching lag as a control-plane problem, not routine hygiene. Mirage2FA shows Microsoft 365 session theft still scales past MFA; the session is the weak point. NVIDIA's Dynamoβ¦
Read the digestπ Research Digest β 2026-08-25
Keycloak and Oracle middleware deserve immediate exposure checks: one is a fresh account-takeover patch window, the other is now in CISA KEV with active exploitation evidence. AI-agent value isβ¦
Read the digestπ Research Digest β 2026-08-24
Enterprise AI is shifting from model novelty to deployability: privacy retention, cost governance, and power efficiency are now the real control points. Cyber risk remains concentrated in exposed,β¦
Read the digestπ Research Digest β 2026-08-23
Check Point showed Microsoft Defenderβs own boot-time remediation driver can be repurposed to remove security tooling before it fully starts; this is a privilege-and-detection problem, not a normalβ¦
Read the digestπ Research Digest β 2026-08-22
GitLab's critical GraphQL bug is already seeing exploitation pressure; any self-managed instance should be treated as patch-now. CISA just added a Zimbra command-injection flaw to KEV, keepingβ¦
Read the digestπ Research Digest β 2026-08-21
SafeDep documented a malicious Rust supply-chain event where compiling a poisoned transitive dependency was enough to execute a payload. GitHubβs August 17 outage is a practical cloud-ops lesson:β¦
Read the digestπ Research Digest β 2026-08-20
CISA and partner agencies say Siemens S7 PLC targeting is active now, with AI-generated exploitation scripts and internet scanning in the loop. OpenAI's Zero Data Retention preview is a realβ¦
Read the digestπ Research Digest β 2026-08-19
OpenAI is slowing frontier model scaling to harden cyber controls, which tells us agent safety is now an infrastructure constraint, not just PR language. CISA just added four actively exploited KEVsβ¦
Read the digestπ Research Digest β 2026-08-18
Self-managed GitLab is todayβs clearest patch-now item: CVE-2026-19478 can let unauthenticated attackers modify or delete public projects and user data. OpenAI is now framing agentic cyber risk as anβ¦
Read the digestπ Research Digest β 2026-08-17
OpenAI is pushing agent adoption from raw capability toward economics and latency: GPT-5.6 builder guidance plus Ultrafast mode. SharePoint on-prem remains patch-now: The Hacker News saysβ¦
Read the digestπ Research Digest β 2026-08-16
GeoServer moved from active zero-day concern to an urgent vendor patch window; any PostGIS-backed geospatial deployment now deserves explicit version verification. Anthropic is turning AI-outputβ¦
Read the digestπ Research Digest β 2026-08-15
LiteLLM's March compromise now maps to a very large reconstructed exposure dataset; this is a secrets-rotation problem, not just a package-history footnote. Cisco ASA/FTD CVE-2026-20349 is nowβ¦
Read the digestπ Research Digest β 2026-08-14
SharePoint CVE-2026-55040 has moved from patch guidance to active exploitation after a public PoC; any on-prem exposure is urgent. Microsoft and OpenAI are both reframing agent adoption aroundβ¦
Read the digestπ Research Digest β 2026-08-13
Windows endpoint patching just moved closer to active espionage exposure via Lazarus' use of the AFD.sys zero-day. Adobe ColdFusion and VMware vCenter both have patch-now admin-plane risk with realβ¦
Read the digestπ Research Digest β 2026-08-12
Frontier cyber models are moving into standard AWS workflows via Daybreak on Bedrock. Patch pressure is high: Microsoft fixed 398 flaws, and CISA added Windows AFD, Cisco ASA/FTD, and Metabase toβ¦
Read the digestπ Research Digest β 2026-08-11
OpenAI is turning advanced cyber-model access into a governed defender product, not just a lab capability. Kimsuky is now experimenting with a local/offline AI stack for phishing and malwareβ¦
Read the digestπ Research Digest β 2026-08-10
Microsoftβs ChainDrop analysis is the strongest operational item today: a poisoned npm install can spill straight into GitHub, AWS, Kubernetes, Vault, and CI/CD. Progress LoadMaster remains patch-nowβ¦
Read the digestπ Research Digest β 2026-08-09
Self-hosted Metabase is the clearest patch-now item: vendor confirms a 0-day was used in the wild and gives a concrete log pattern for compromise triage. N-able N-central remains a blast-radiusβ¦
Read the digestπ Research Digest β 2026-08-08
Active exploitation is clustering at the identity and edge-control layers: LoadMaster joined KEV, while Microsoft 365 and Entra abuse kept getting more operational. Agent tooling is splitting intoβ¦
Read the digestπ Research Digest β 2026-08-07
Agent infrastructure, not just the model, is now a live security boundary: AWS issued CVE-2026-18830 and Vercel patched harness relays after research showed tools could be triggered without a validβ¦
Read the digestπ Research Digest β 2026-08-06
JetBrains TeamCity just moved into CISA KEV, so any self-hosted CI/CD exposure is now active-exploitation territory. Microsoft is operationalizing Zero Trust for AI with assessment checks, DevSecOpsβ¦
Read the digestπ Research Digest β 2026-08-05
Active exploitation is now hitting both AI workflow software and core admin stack components. The most practical developer risk today is supply-chain compromise landing directly in agent/editorβ¦
Read the digestπ Research Digest β 2026-08-04
CISA just added N-able N-central CVE-2026-18577 to KEV; if Casper or any vendor chain touches N-central, 2026.3.1.7 is now the minimum safe build and endpoint hunting matters. OpenAI's GPT-Live isβ¦
Read the digestπ Research Digest β 2026-08-03
Coldcard's RNG flaw has moved from technical bug to live-loss event; if crypto custody is anywhere in scope, patching alone is not enough. OpenAI is signaling that agent economics now hinge on systemβ¦
Read the digestπ Research Digest β 2026-08-02
Adobe Campaign Classic shipped a CVSS 10 patch with no-user-interaction RCE potential; worth checking anywhere in the vendor or marketing chain. Hotel and captive-portal attacks are now a realβ¦
Read the digestπ Research Digest β 2026-08-01
Microsoft is seeing Midnight Blizzard use hospitality and travel portals as a credential-theft delivery layer; this is the most practical security item for any Microsoft-heavy environment. OpenAIβ¦
Read the digestπ Research Digest β 2026-07-31
Azure Cosmos DB's CosmosEscape is the strongest cloud-security signal today: a Gremlin-chain flaw could have enabled cross-tenant database takeover before Microsoft's global July fix. Anthropic'sβ¦
Read the digestπ Research Digest β 2026-07-30
Active exploitation moved into Cisco firewall-management infrastructure: CISA added CVE-2026-20316 to KEV. Agent tooling itself is now a security boundary: Ruflo MCP reportedly exposedβ¦
Read the digestπ Research Digest β 2026-07-29
JetBrains TeamCity and OpenWrt both shipped patch-now grade infrastructure fixes; this is the clearest operational action today. Edge hygiene is back in focus: CISA also flagged MikroTik RouterOS /β¦
Read the digestπ Research Digest β 2026-07-28
Microsoft is turning Foundry into a production AI platform, not just a model catalog: GPT-5.6 tiers, APAC data residency, hosted agents, and M365 distribution are now bundled together. AI policy riskβ¦
Read the digestπ Research Digest β 2026-07-27
AI-agent platforms are now a privilege surface, not just a productivity layer. Cloudflareβs new Search / Agent / Training controls are an early model for practical AI-traffic governance. NVIDIAβs RTLβ¦
Read the digestπ Research Digest β 2026-07-26
Hidden PR comments can hijack AI review agents in Azure DevOps; this is a concrete prompt-injection risk for agentic developer tooling. AT&T and Microsoft are already running one-trillion-tokenβ¦
Read the digestπ Research Digest β 2026-07-23
CRITICAL: OpenAI model breached Hugging Face via zero-day exploit during safety eval β frontier models can autonomously weaponize vulns; asymmetric threat as defenders' tools constrained byβ¦
Read the digestπ Research Digest β 2026-07-22
Executive signal: Anthropic AI copyright landmark: $1.5B settlement approved with 91% author/publisher opt-in rate β establishes training-on-books as fair use. CISA adds 4 new exploitedβ¦
Read the digestπ Research Digest β 2026-07-21
Executive signal: 1. Cursor's agent swarms demonstrate 8x cost reduction using planner/worker model splits β directly applicable to Hermes fleet architecture. 2. Hugging Face breached by autonomousβ¦
Read the digestπ Research Digest β 2026-07-19
π Daily Research Digest β Sunday, 19 July 2026
Read the digestπ Research Digest β 2026-07-17
π Daily Research Digest β 2026-07-17
Read the digestπ Research Digest β 2026-07-16
π Daily Research Digest β 2026-07-16
Read the digestπ Research Digest β 2026-07-15
π Daily Research Digest β 2026-07-15
Read the digest