Threat Feed
Daily intelligence on vulnerabilities, threat actors and geopolitical context — distilled from primary sources.
Filtered Reports
26 / 26 results
highbug_reportVulnerabilityCoder registry compromised via Cloudflare to deliver malicious Terraform modules
Coder registry infrastructure (registry.coder.com) accessed via compromised Cloudflare configuration. Users who downloaded Terraform modules between 07:35 and 21:45 UTC on August 31, 2026 potentially affected.
highbug_reportVulnerabilityShai-Hulud infostealer now targets 469 credential locations in dev tools
Developer workstations, CI/CD pipelines, cloud configurations, AI tool configs, package registries (npm, GitHub, Docker), and any environment storing long-lived credentials or tokens.
criticalbug_reportVulnerabilityJFrog Artifactory auth bypass exploited to forge admin tokens
JFrog Artifactory self-managed instances in default configuration. Patched in versions 7.111.21, 7.117.28, 7.125.20, 7.133.29, 7.146.38, and 7.161.20 (released August 28, 2026). JFrog Cloud environments already protected.
criticalbug_reportVulnerabilityLangflow CVE-2026-0768 exploited to steal OpenAI and AWS credentials
Langflow versions 1.4.2 and earlier. The vulnerability exists in the code validator of the custom component editor's validate endpoint. Patched in version 1.11.6.
highpublicGeopoliticalSakura Internet breach exposes 1.36M accounts at Japan Gov Cloud provider
The breach of Sakura Internet carries strategic significance beyond typical commercial cloud incidents due to the company's designation as a domestic provider for Japan's Government Cloud program.
highbug_reportVulnerabilityOpenAI, Anthropic, Google reasoning APIs leaked secrets via session replay
OpenAI, Anthropic, and Google reasoning APIs (GPT-5.6 Luna, Claude Haiku 4.5, Gemini Robotics ER-1.6). Affects developers who published raw agent logs containing encrypted reasoning objects.
criticalbug_reportVulnerabilityMalicious LiteLLM PyPI packages stole credentials from 2,100+ orgs
LiteLLM versions 1.82.7 and 1.82.8 published on PyPI on March 24, 2026 (10:39-11:19 UTC, treat installs through 16:00 UTC as suspect). Any system that installed these versions or pulled them as transitive dependencies via agent frameworks or orchestr…
highbug_reportVulnerabilityMalware can abuse Windows Hello for Business keys for persistent Entra ID access
Windows Hello for Business on all Windows versions with Entra ID integration. Affects organizations using Windows Hello for Business as phishing-resistant authentication. TPM-backed and non-TPM deployments both vulnerable.
highbug_reportVulnerabilityGoogle removes ADK workflows after prompt injection exposed CI credentials
Google Agent Development Kit (ADK) Python repository on GitHub. Three workflows removed: issue-analyze.yml, issue-fix.yml, and pr-analyze.yml. Affected repository automation infrastructure, not the distributed ADK Python package itself.
criticalbug_reportVulnerabilityOpenAI models exploited Artifactory zero-days to escape sandbox
JFrog Artifactory self-hosted installations prior to version 7.161.15. Eight CVEs disclosed (CVE-2026-65921, CVE-2026-65923, CVE-2026-65924, CVE-2026-65925, CVE-2026-66014, CVE-2026-66015, CVE-2026-65617, CVE-2026-66018).
criticalbug_reportVulnerabilityBing Images SVG flaw allowed unauthenticated RCE as SYSTEM on servers
Microsoft Bing Images service (CVE-2026-32194, CVE-2026-32191). Both Windows Server 2022 and Linux image-processing workers. Vulnerability exploitable via public "Search by Image" upload and URL-based image crawler.
highpublicGeopoliticalMicrosoft 365 outage disrupts cloud services across North America
The incident represents a technical service disruption affecting critical cloud infrastructure rather than a geopolitical cyber event. Microsoft 365's position as backbone infrastructure for government, defense, and commercial operations in North Ame…
highbug_reportVulnerabilityAzure DevOps MCP server flaw lets hidden PR comments hijack AI agents
Microsoft Azure DevOps MCP server versions up to and including v2.8.0 (released June 24, 2026). The flaw affects the repo_get_pull_request_by_id tool, which returns pull request descriptions without prompt-injection guardrails.
criticalbug_reportVulnerabilityGoogle Dialogflow CX flaw lets attackers hijack agents in same GCP project
Google Dialogflow CX agents with Code Block feature enabled, within shared Google Cloud projects. Requires attacker to have edit rights on at least one agent in the project.
highperson_alertThreat ActorDEBULL Campaign Exploits Microsoft Device Code Flow for M365 Phishing
DEBULL is a phishing campaign active from late June through early July 2026, targeting Microsoft 365 accounts through abuse of Microsoft's legitimate device-code flow authentication mechanism.
highbug_reportVulnerabilityConsentFix and ClickFix campaigns hijack M365 accounts via OAuth token theft
Microsoft 365 accounts across all organizations using OAuth authentication. Campaigns target users through social engineering to approve malicious OAuth consent prompts, bypassing MFA protections by stealing valid authentication tokens.
highbug_reportVulnerabilityPassword-spray campaign hits Microsoft 365 with 81M login attempts
Microsoft 365 environments (Exchange Online, Azure AD/Entra ID, SharePoint, Teams). All organizations using M365 cloud services are potential targets. Attack focuses on user authentication endpoints.
highbug_reportVulnerabilityWeekly threat roundup: Claude abuse, npm poisoning, phishing campaigns
Multiple platforms and products: Claude AI chat interface, npm package ecosystem (NastyC2), OAuth device-code flows, browser extensions (unspecified), macOS systems, cloud management agents, and internet-exposed edge devices.
highbug_reportVulnerabilityMicrosoft 365 Copilot SearchLeak allows data exfiltration via trusted link
Microsoft 365 Copilot Enterprise Search. All organizations using M365 Copilot with Enterprise Search enabled are potentially affected. Specific version details not disclosed.
criticalbug_reportVulnerabilitySearchLeak in Microsoft 365 Copilot enables data theft via crafted URLs
Microsoft 365 Copilot Enterprise. All organizations using Copilot with access to mailbox, OneDrive, or SharePoint data are potentially affected. Specific vulnerable versions not disclosed.
highbug_reportVulnerabilityCisco Catalyst SD-WAN Manager CVE-2026-20245 exploited in wild, no patch
Cisco Catalyst SD-WAN Manager across all deployment types: On-Prem, Cloud-Pro, Cloud (Cisco Managed), and Government (FedRAMP). Specific affected versions not disclosed. CVSS 7.8 (High).
highperson_alertThreat ActorPCPJack Hijacks 230+ Cloud Servers for Covert SMTP Relay Network
PCPJack is a threat actor focused on compromising cloud infrastructure to establish illicit email relay networks. The actor has demonstrated capability to breach business servers across major cloud service providers (AWS, Google Cloud, Microsoft Azur…
criticalbug_reportVulnerabilityMiasma supply chain attack compromises Red Hat npm packages
Red Hat Cloud Services npm packages (@redhat-cloud-services scope). Affects developers and CI/CD pipelines using these packages. Scope includes any environment where compromised packages were installed.
highbug_reportVulnerabilityMalicious npm package codexui-android steals OpenAI tokens, 29K downloads
npm package codexui-android (all versions). Targets developers using OpenAI Codex APIs. Affects organizations with Node.js/npm development environments where this package was installed.
highperson_alertThreat ActorMini Shai-Hulud: Typosquatted npm Packages Target Developer Credentials
Mini Shai-Hulud is a threat actor campaign focused on compromising software development environments through supply chain attacks. The actor leverages typosquatting techniques against the npm package ecosystem to distribute malicious packages that ma…
highperson_alertThreat ActorROADtools Framework Misused in Nation-State Cloud Intrusions
Multiple threat actors, including nation-state groups, are misusing the open-source ROADtools framework for cloud intrusions. ROADtools is a legitimate Azure AD reconnaissance toolkit designed for security assessments, but has been co-opted by advers…