12 stories · 2 sources · 3 critical · 4 high · ~14 min read

Coverage: Last 72 hours

Today’s Highlights

This cycle highlights widespread attacker innovation: from exploitation of trust in browser extensions to critical authentication and supply chain weaknesses, plus the growing operational complexity of defending AI-driven environments. Attackers are leveraging social engineering to bypass passwordless protections on Microsoft accounts and exploiting critical vulnerabilities in platforms such as GitLab, Artifactory, and cloud routers. Meanwhile, large-scale credential theft, SOC alert fatigue, and AI-powered attacks are reshaping both the threat landscape and defensive priorities.

Table of Contents

1. Telus Warns Customers of Account Breaches
2. CISA Adds 5 Actively Exploited Artifactory, ScreenConnect, and RouterOS Flaws to KEV
3. Claude Used to Automate Exploitation and Data Theft Across Multiple Victims
4. GitLab CVSS 10 File-Read Flaw Draws In-the-Wild Probes After Disclosure
5. Malicious Twitch Browser Extension Leaks OAuth Tokens From Nearly 31,000 Users
6. Attackers Use Passkey Phishing to Hijack Microsoft Cloud Accounts and Exfiltrate Data
7. Russian State-Sponsored Hackers Use Claude to Rebuild Malware After Detection
8. Your Critical Vulnerabilities Might Not Be Your Biggest Risk

Critical   High   Medium   Low

Top Stories

Telus Warns Customers of Account Breaches

Source: SecurityWeek | Published: Sep 14 | Risk: HIGH | Impacted: Telus subscribers, Canadian telecom customers | Topics: Data Breach / Threat Intel

What happened: Stolen credentials were used in a multi-month campaign to access subscriber personal data and billing records. The post Telus Warns Customers of Account Breaches appeared first on SecurityWeek.

Why it matters: The compromise of Telus subscriber accounts via stolen credentials exposes a broad set of sensitive information over an extended timeframe, enabling potential fraud, privacy breach, and data resale in criminal markets.

How it works: Attackers maintained sustained access by using stolen login details to bypass standard authentication and extract billing and personal information, likely through automated credential stuffing or targeted phishing attacks.

Practitioner Perspective

Telecom companies hold a wealth of sensitive customer data and are frequent targets for credential reuse campaigns. Organizations relying on Telus services should expect downstream risk to single sign-on, phone number-based authentication, or customer support workflows that may involve compromised user data. Defensive teams need to prepare for fraud escalations, both from compromised account holders and from threat actors using breached information to target additional services.

Recommended Actions

• Force password resets and enable mandatory multi-factor authentication for affected user segments in the Telus customer base
• Flag and monitor for suspicious logins or changes in high-value customer accounts using risk-based authentication

Exploits & CVEs

CISA Adds 5 Actively Exploited Artifactory, ScreenConnect, and RouterOS Flaws to KEV

Source: The Hacker News | Published: Sep 12 | Risk: CRITICAL | Impacted: DevOps pipelines, Remote access environments, Edge routers/gateways | Topics: Exploit / Vulnerability

What happened: CISA added five actively exploited vulnerabilities to its Known Exploited Vulnerabilities catalog: two in JFrog Artifactory, one in ConnectWise ScreenConnect, and two in MikroTik RouterOS. These flaws could lead to privilege escalation, unauthorized file transfers, and remote code execution. Agencies are urged to apply patches by specified deadlines.

Why it matters: Active exploitation of unpatched Artifactory, ScreenConnect, and RouterOS vulnerabilities creates real-world risk of remote code execution and privilege escalation on supply chain and edge infrastructure components.

How it works: Artifactory is a binary repository manager. ScreenConnect is remote access software. MikroTik RouterOS runs on internet-facing network hardware. These vulnerabilities allow remote attackers to execute code or escalate privileges if devices are not patched.

Affected / Fix: Patch deadlines and affected product names listed in CISA KEV; apply vendor advisories for identified CVEs.

Practitioner Perspective

Any environment running JFrog Artifactory, ConnectWise ScreenConnect, or MikroTik RouterOS must treat these CVEs as actively in use by attackers. Exploitation can bypass segmentation controls if endpoints interface with build systems, source repositories, or remote site routers. Security teams should escalate patching these products to emergency status, not just routine patch windows. Missing these patches could lead to attacker persistence inside CI/CD workflows or lateral movement from compromised edge devices. Move quickly: assume proof-of-concept exploit code is circulating in threat actor channels.

Recommended Actions

• Apply vendor patches for the five CVEs in JFrog Artifactory, ScreenConnect, and MikroTik RouterOS flagged by CISA KEV
• Review security group and firewall policies for unneeded exposure of these products to the internet

Claude Used to Automate Exploitation and Data Theft Across Multiple Victims

Source: The Hacker News | Published: Sep 11 | Risk: CRITICAL | Impacted: Enterprises with malware-impacted endpoints, Antivirus vendors, Critical infrastructure SOCs | Topics: Exploit / Vulnerability

What happened: Anthropic has warned that cybercriminals and state-sponsored hackers alike are using its Claude models for cyber attacks, weapons design, propaganda, and mass surveillance between December 2025 and August 2026. The threat actors, which the artificial intelligence (AI) company has branded Generative Threat Groups (GTGs), span state-sponsored groups, financially motivated criminals, commercial

Why it matters: The use of powerful commercial AI models like Claude to automate tasks ranging from exploitation to data theft enables attackers to dramatically increase capability and scale, bypassing many traditional detection and prevention measures with ease.

How it works: Attackers leverage AI to script, modify, and deploy attacks at speed, automating processes for exploitation or surveillance that previously required bespoke malware development and manual testing.

Practitioner Perspective

The weaponization of models like Claude marks a turning point in adversarial capability, empowering even less sophisticated groups to execute advanced tactics using generative AI for rapid tooling and operational flexibility. SOC and infrastructure defenders must assume higher levels of campaign automation and obfuscation, requiring adaptive defensive techniques and threat modeling for novel attack applications.

Recommended Actions

• Integrate threat intelligence feeds for AI-abuse and Generative Threat Group indicators of compromise (IOCs) into SIEM and analytics platforms
• Prioritize advanced behavioral and process-based detections over signature-based methods for likely AI-driven threats

Emerging Signals

GitLab CVSS 10 File-Read Flaw Draws In-the-Wild Probes After Disclosure

Source: The Hacker News | Published: Sep 11 | Risk: CRITICAL | Impacted: Organizations running self-hosted GitLab, CI/CD infrastructure owners, DevOps teams | Topics: Exploit / Vulnerability

What happened: GitLab addressed a critical CVSS 10.0 vulnerability (CVE-2026-85706) in its repository commits API, allowing unauthenticated users to read arbitrary server files. Active exploitation was detected within hours of disclosure, prompting immediate patching.

Why it matters: An unauthenticated file read vulnerability in GitLab exposes all sensitive configuration and secrets stored on affected servers if left unpatched, creating opportunity for supply chain compromise and privilege escalation.

How it works: GitLab is a popular self-hosted DevOps platform. CVE-2026-85706 (CVSS 10.0) allows unauthenticated users to read arbitrary files on the server by abusing the repository commits API, exposing sensitive data without needing to log in.

Affected / Fix: Patch available for CVE-2026-85706; organizations should apply immediately per vendor advisory.

Practitioner Perspective

Security teams should assume that attackers are actively probing GitLab instances as soon as these types of CVEs are announced. The file read weakness means attackers can discover secrets, keys, and other sensitive server files without credentials, which can then be used for lateral movement or code signing attacks down the CI/CD pipeline. Maintaining unpatched GitLab servers with public exposure creates an unacceptable risk of business compromise by both cybercriminals and APT groups. Prioritize rapid patching and review segmented access if patching is delayed. Defenders should monitor for reconnaissance activity against GitLab endpoints and treat unexplained project or repository changes with high suspicion.

Recommended Actions

• Deploy the latest GitLab patch addressing CVE-2026-85706 to all affected instances immediately
• Restrict external access to GitLab repository commit APIs until confirmation of patch deployment

Malicious Twitch Browser Extension Leaks OAuth Tokens From Nearly 31,000 Users

Source: The Hacker News | Published: Sep 14 | Risk: HIGH | Impacted: Twitch users, Browser extension users, SaaS identity environments | Topics: Identity / Exploit

What happened: A malicious Twitch browser extension named ‘Twitch Enhanced Viewer | JeetBot’ leaked OAuth tokens from nearly 31,000 users to proxy servers operated by a Russian bot service. The extension, available on Chrome and Firefox, forwarded users’ OAuth tokens in cleartext, potentially exposing their Twitch accounts to unauthorized access.

Why it matters: OAuth token leakage through malicious browser extensions can grant attackers persistent access to user accounts, bypassing most standard authentication controls and making forensic containment difficult.

How it works: OAuth is an authentication protocol widely used for delegated access to cloud accounts. When browser extensions exfiltrate OAuth tokens in cleartext, attackers can use them to impersonate users and directly access protected resources.

Practitioner Perspective

Organizations with workforces or communities using Twitch should assume a percentage of user accounts have been compromised if this extension was present. Browser extension risk extends well beyond network boundaries: attackers can silently harvest session tokens for abuse or impersonation. In regulated sectors, external account takeovers can become incident-reporting events if tied to protected data or business activity. Security teams should inventory and restrict unapproved extensions and prioritize revocation of exposed credentials. The key concern is how silently credential theft like this can pivot to fraud or further lateral movement.

Recommended Actions

• Revoke and re-issue all OAuth tokens for users identified with ‘Twitch Enhanced Viewer | JeetBot’ installed
• Block or remove the ‘Twitch Enhanced Viewer | JeetBot’ extension on Chrome and Firefox via enterprise browser management tools

Attackers Use Passkey Phishing to Hijack Microsoft Cloud Accounts and Exfiltrate Data

Source: The Hacker News | Published: Sep 13 | Risk: HIGH | Impacted: Microsoft 365 tenants, Azure AD/Entra ID organizations, Cloud administrators | Topics: Exploit / Vulnerability

What happened: Cybercriminals are exploiting passkey-themed social engineering tactics to infiltrate Microsoft cloud accounts. By impersonating IT help desk personnel, they convince users to update their passkeys or multi-factor authentication settings, leading to unauthorized access and data exfiltration.

Why it matters: Impersonation and phishing campaigns targeting passkey and MFA reset processes allow attackers to undermine core identity trust for cloud accounts, enabling data theft and privilege escalation even in mature organizations.

How it works: Passkeys are FIDO2-based passwordless credentials. Attackers leverage social engineering to trick users into enrolling attacker-controlled passkeys or altering multi-factor authentication registrations, granting unauthorized account access.

Practitioner Perspective

Organizations relying on Microsoft identity and cloud services are prime targets for social engineering campaigns abusing passkey and MFA migration confusion. Attackers understand users’ limited familiarity with passkey workflows, exploiting help desk impersonation to capture new authentication material. This sidesteps traditional phishing-resistant methods if users can be lured into registering malicious or attacker-controlled passkeys. Security teams need to harden identity enrollment and change processes so that attackers cannot easily exploit passwordless technologies. The top priority is breaking attacker feedback cycles around social engineering in authentication management.

Recommended Actions

• Audit all recent passkey and MFA resets in Microsoft cloud environments for anomalous help desk activity
• Strengthen out-of-band verification for MFA or passkey enrollment requests, especially those initiated remotely

Russian State-Sponsored Hackers Use Claude to Rebuild Malware After Detection

Source: The Hacker News | Published: Sep 11 | Risk: HIGH | Impacted: Enterprises with malware-impacted endpoints, Antivirus vendors, Critical infrastructure SOCs | Topics: Exploit / Vulnerability

What happened: Anthropic on Thursday revealed it disrupted a campaign mounted by a Russian state-sponsored threat actor that abused Claude for developing an AI-assisted workflow to get ahead of the detection curve. The operation has been attributed to a cyber espionage group it calls GTG-20006 (where “GTG” stands for Generative Threat Group), which aligns with broader reporting linking the cluster to Midnight

Why it matters: State-aligned threat actors leveraging commercial AI models for rapid malware adaptation undermine traditional signature-based detection and dramatically compress defenders’ reaction windows.

How it works: Commercial AI models such as Claude can be used by attackers not just for direct exploitation but also to automate rewriting or recompiling malware when detected, enabling faster evasion cycles.

Practitioner Perspective

This incident shows that sophisticated adversaries are now operationalizing AI like Claude to defeat even recently-updated malware detection tools. Each detection turns into a feedback loop for attackers to rework dangerous functionality, raising the bar for defense across both enterprise and critical infrastructure sectors. Defensive teams must not only upgrade detection methodologies but also monitor for signs of automated malware devops on endpoints. Expect the cost and volume of evasive malware to rise as AI workflows become standard in threat actor toolkits. The core concern: defenders face automation arms races that favor rapid attacker iteration.

Recommended Actions

• Monitor for indicators of Claude or other commercial AI model usage in malware variant development workflows
• Enhance endpoint detection rules for non-human process chains and rapid malware retooling patterns

Your Critical Vulnerabilities Might Not Be Your Biggest Risk

Source: The Hacker News | Published: Sep 11 | Risk: MEDIUM | Impacted: Vulnerability management teams, ICS/OT operators, Network architects | Topics: Exploit / Vulnerability

What happened: Security teams have become exceptionally talented at finding vulnerabilities. Now, it’s time to turn our attention to optimizing the process for determining which of those vulnerabilities actually create a path to compromise. A critical vulnerability may look alarming on a scanner report, but if it sits behind strong segmentation, identity controls, and other defenses that prevent an attacker

Why it matters: Critical CVEs can present minimal real risk if compensating controls block exploitation paths, while low-severity findings may offer attackers direct paths to compromise if unmitigated.

How it works: Vulnerability scanners produce technical findings with risk scores (CVSS), but exploitation depends on the broader security architecture, attackers look for actual pathways, not just disclosed exposures.

Practitioner Perspective

Vulnerability management programs routinely over-prioritize scanner scores at the expense of actual exploitability. Segmentation, least-privilege, and incident response readiness often reduce the impact or likelihood of critical findings, but few security teams correlate compensating control strength against scanner output. Defenders should triage vulnerabilities by true attacker paths, including identity, network posture, and business process context, rather than defaulting to CVSS alone. Focusing on business-impacting attack chains, not just CVE ratings, is the goal. The real risk: prioritizing optics over adversarial realism undermines true resilience.

Recommended Actions

• Map critical CVEs to compensating controls such as segmentation and strong authentication in risk assessment workflows
• Deprioritize CVE remediation tickets backed by high-confidence, attacker-blocking controls

Also Today

• When the Whole Company Adopts AI: What It Does to Your SOC: AI-related alerts are rising rapidly in Security Operations Centers, with most being false positives but true positive alert volume is expected to increase.
• OpenAI Agents Linked to RubyGems Campaign That Gained RCE on RubyDoc Servers: OpenAI agents were used to upload over 2,000 malicious packages to RubyGems, exploiting RubyDoc.info build flaws for data exfiltration.
• Anthropic Says Seven China-Based AI Labs Ran Industrial-Scale Claude Distillation Attacks: Anthropic identified and disrupted large-scale model distillation campaigns by Chinese labs targeting its Claude platform.
• CISOs Race to Control AI Agents Without Destroying Their Value: Security leaders are seeking to modernize their control over AI agents to prevent over-privilege while maintaining business value.

Defensive Actions

• Revoke and re-issue all OAuth tokens for accounts affected by the ‘Twitch Enhanced Viewer | JeetBot’ extension
• Block or remove malicious browser extensions in Chrome and Firefox using enterprise management tools
• Apply urgent vendor patches for currently exploited CVEs in JFrog Artifactory, ScreenConnect, MikroTik RouterOS, and GitLab
• Audit and monitor passkey and multi-factor authentication (MFA) reset events in Microsoft cloud environments for anomalies
• Harden out-of-band verification for all cloud identity changes and privilege escalations, especially for remote-initiated requests
• Update SOC procedures and SIEM suppression rules to more effectively triage AI-related security alerts and reduce analyst noise
• Map critical vulnerabilities to actual attacker pathways, adjusting remediation priorities based on compensating controls and network segmentation

What We’re Watching

• Active exploitation of GitLab CVE-2026-85706 (CVSS 10.0): monitor for post-patch threat actor activity
• Continued passkey and MFA phishing against Microsoft 365 and Azure Entra ID tenants
• Proof-of-concept and weaponized exploits for CISA KEV-listed CVEs in Artifactory, ScreenConnect, and MikroTik RouterOS
• AI-powered malware retooling by Generative Threat Groups, especially state-linked actors using Claude
• RubyGems, npm, and PyPI repositories for new large-batch attempts at supply chain compromise using AI-assisted code generation

Found this briefing useful? Follow the blog to get the next one as soon as it is published, and pass it along to a colleague who owns patching.

Categories: Cybersecurity Blog, Cybersecurity News

Tags: AI, Cloud, cve, cybersecurity, daily digest, Identity, Phishing