Daily Brief

Forrester projects AI and automation could eliminate 10.4 million US jobs by 2030, equating to 6.1% of the current workforce. The anticipated job losses are considered structural and permanent, contrasting with cyclical losses experienced during economic recessions. AI is expected to augment rather than replace roles, with one in five positions potentially impacted, necessitating investment in staff training. Some SaaS providers, including Salesforce and Workday, are already integrating AI to replace parts of their workforce. Over-automation risks include potential reputational damage and weakened employee experiences, as noted by companies like Duolingo and Klarna. Forrester warns that many layoffs attributed to AI are financially driven, with AI often used as a scapegoat. The rise of agentic AI and generative AI projects has shifted the forecast, with 50% of automation-related job losses now linked to these technologies. Despite these changes, Forrester predicts that humans will continue to perform most work tasks over the next five years.

Betterment, a leading U.S. digital investment advisor, experienced a data breach, leading to fraudulent crypto-related emails being sent to a subset of its customers. The breach involved unauthorized access to a third-party marketing platform, not Betterment's core systems, ensuring customer accounts and credentials remained secure. Hackers used the compromised system to send emails from a legitimate Betterment subdomain, falsely promising to triple cryptocurrency deposits. The company promptly warned customers about the scam, confirming the fraudulent nature of the messages and advising caution against unexpected communications. Betterment is enhancing defenses against social engineering attacks and plans to release a detailed incident analysis once investigations conclude. This incident mirrors a similar attack on Grubhub, suggesting the potential reuse of tactics by the same threat actor. Betterment manages over $65 billion in assets for more than one million customers, underscoring the importance of robust cybersecurity measures in the financial sector.

A new phishing tactic on LinkedIn involves fake "reply" comments that mimic official LinkedIn branding, misleading users into visiting external links under the guise of policy violations. The scam employs LinkedIn's own URL shortener, lnkd.in, to disguise phishing links, making them difficult to distinguish from legitimate URLs, especially on certain devices. Fraudulent comments claim users' accounts are "temporarily restricted" due to non-compliance, prompting them to verify their identity through malicious links. Phishing sites like very1929412.netlify[.]app are used to harvest credentials by further misleading users to verify their accounts. Fake company pages, such as "Linked Very," are being used to post these deceptive comments, exploiting LinkedIn's official logo and branding. LinkedIn has acknowledged the issue and is actively working to dismantle the campaign, urging users to report suspicious activity for prompt action. Users are advised to remain vigilant and avoid interacting with comments or messages that impersonate LinkedIn, especially those urging clicks on external links.

Dutch police apprehended a 33-year-old suspect at Schiphol Airport, believed to be behind the AVCheck online platform used by cybercriminals. The arrest follows an international investigation involving authorities from the Netherlands, United States, and Finland, part of Operation Endgame. AVCheck was a service enabling cybercriminals to test malware against antivirus products, crucial for evading detection in cyberattacks. The suspect had deregistered in the Netherlands and moved to the UAE around the time AVCheck was dismantled in May 2025. Authorities seized data storage devices from the suspect, potentially containing critical evidence for ongoing investigations. AVCheck's takedown marks a significant blow to the cybercrime ecosystem, disrupting a major tool for refining and deploying malware. This operation demonstrates the effectiveness of international cooperation in combating sophisticated cybercriminal networks.

The rise of agentic AI tools like Copilot and Codex is transforming software engineering, enabling rapid end-to-end software development but introducing significant security challenges. Machine Control Protocols (MCPs) are crucial yet often unsecured layers that dictate AI agent capabilities, including tool access and API interactions. A recent vulnerability, CVE-2025-6514, exploited a trusted OAuth proxy, affecting over 500,000 developers, demonstrating the potential for AI-driven remote code execution. This incident underscores the need for robust control over AI agents, as compromised MCPs can lead to authoritative and potentially malicious actions by AI systems. The webinar aims to educate security teams on the inherent risks of agentic AI, focusing on MCP server operations, shadow API key proliferation, and permission sprawl. Traditional identity and access models face challenges in managing AI agents, highlighting the necessity for updated security frameworks to prevent unauthorized actions. Participants will gain insights into identifying and mitigating AI-related vulnerabilities to maintain control over their AI infrastructure and prevent future incidents.

Target's internal source code and documentation were leaked, confirmed by current and former employees, raising concerns about data security and potential exposure. The breach involved the unauthorized sale of Target's source code on Gitea, a public software development platform, with a sample confirming the authenticity of the leaked data. Target implemented an "accelerated" security measure, restricting access to its Enterprise Git server to internal networks or VPNs following the breach notification. The root cause of the breach remains undetermined, though infostealer malware compromised a Target employee's workstation, potentially contributing to the data leak. The leaked dataset is reportedly 860GB, though only a 14MB sample has been reviewed, containing authentic internal code and system references. Target has not publicly addressed the breach or potential insider involvement, despite being informed of the situation and offered assistance for investigation. This incident underscores the importance of robust access controls and monitoring to prevent unauthorized data access and potential insider threats.

CISA has mandated federal agencies to address a critical vulnerability in the Gogs Git server, following its addition to the Known Exploited Vulnerabilities catalog. The flaw, identified as CVE-2025-8110, allows authenticated users to execute remote code by exploiting a path traversal vulnerability, posing significant risks. Over 700 Gogs instances have already been compromised, with more than 1,400 servers potentially vulnerable, highlighting the urgency of the situation. Agencies are instructed to either implement mitigations or cease using Gogs if no viable workarounds exist, to prevent further exploitation. The vulnerability was initially discovered by Wiz researchers, who noted it bypassed previous fixes due to unaccounted symbolic links. Gogs users are advised to disable open registration and employ VPNs as temporary measures until a permanent fix is available. While the attackers remain unidentified, the use of Supershell C2 suggests potential links to actors based in Asia. This incident serves as a reminder of the critical need for robust patch management and proactive vulnerability assessments.

Mandiant has launched AuraInspector, an open-source tool designed to detect and mitigate misconfigurations in Salesforce Aura, enhancing data security for Salesforce admins. The tool automates the identification of common access control issues within Salesforce Experience Cloud sites, focusing on potential data exposure risks. Misconfigurations in Salesforce Aura can allow unauthorized access to sensitive data, such as exploiting the getItems method in Salesforce Account objects. Attackers can bypass the 2,000-record limit per request by altering sort orders or abusing the default GraphQL API access, potentially exposing large data sets. AuraInspector is read-only, ensuring it does not alter Salesforce instances, but rather provides remediation strategies to address security gaps. Despite the shift to Lightning Web Components, Aura remains prevalent in legacy systems, continuing to present security challenges for organizations. Previous incidents highlighted by Varonis and Brian Krebs have shown that misconfigured Salesforce sites can lead to significant data leaks, affecting sectors like banking and healthcare.

A Dutch appeals court upheld a seven-year sentence for a hacker who breached port IT systems using malware-laden USB sticks to aid cocaine smuggling operations. The defendant's appeal claimed police improperly accessed encrypted chats; however, the court found the argument insufficient to overturn the conviction. The hacker orchestrated the breach by convincing a terminal employee to insert a malware-infected USB, allowing remote access for several months. Encrypted SkyECC chats revealed the hacker's detailed involvement, including attempts to gain admin access and discussions on using a hardware keylogger. The court dismissed claims that the hack was lawful due to employee cooperation, emphasizing unauthorized access to sensitive data. The breach facilitated the smuggling of 210 kg of cocaine, hidden in wine shipments with falsified documents and precise logistical coordination. While one drug charge was dropped, the court ordered the hacker to pay for cleanup and legal expenses, and confiscated his hacking equipment.

Check Point Research has identified VoidLink, a sophisticated malware framework targeting Linux-based cloud and container environments, first discovered in December 2025. VoidLink's modular design includes over 30 plugin modules, allowing operators to alter capabilities and adapt to changing objectives, inspired by Cobalt Strike's approach. The malware is attributed to China-affiliated threat actors, focusing on cloud services like AWS, Google Cloud, and Microsoft Azure, and adapting to Docker and Kubernetes environments. VoidLink can gather credentials from cloud systems and version control platforms such as Git, indicating potential targets include software developers for data theft or supply chain attacks. The framework includes a Chinese web-based dashboard for remote control, enabling attackers to manage files, tasks, and plugins, and execute comprehensive attack cycles. VoidLink features anti-analysis capabilities, including self-modifying code and risk-based evasion strategies, allowing it to bypass detection and adapt to various security environments. Developers of VoidLink demonstrate advanced technical skills in multiple programming languages and operating system internals, creating a robust post-exploitation framework.

Cybercriminals in 2025 are optimizing traditional attack methods with AI, focusing on supply chain vulnerabilities and phishing, rather than adopting entirely new strategies. The Shai Hulud NPM campaign illustrates how a single compromised package can impact thousands of downstream projects, highlighting persistent supply chain risks. AI has lowered the barrier to entry for cybercrime, enabling smaller teams or individuals to execute sophisticated attacks that previously required larger operations. Phishing remains effective due to human error, as demonstrated by a recent npm supply chain attack that spread widely after a developer's credentials were compromised. Official platforms, such as Chrome extensions, continue to face challenges in preventing malware, with attackers exploiting insufficient permission models to bypass security measures. The article suggests prioritizing improvements in permissions models, supply chain verification, and phishing-resistant authentication to counteract these optimized traditional tactics. A webinar by OX is announced to discuss current attack techniques and effective defense strategies, emphasizing the importance of reinforcing cybersecurity fundamentals.

ServiceNow has patched a critical vulnerability in its AI Platform, CVE-2025-12420, which allowed unauthenticated user impersonation with a CVSS score of 9.3. The flaw enabled attackers to perform operations as an impersonated user, posing significant security risks to affected organizations. ServiceNow released a security update on October 30, 2025, covering most hosted instances, and distributed patches to partners and self-hosted clients. Aaron Costello from AppOmni discovered and reported the vulnerability, leading to a swift response from ServiceNow to mitigate potential threats. No active exploitation of the vulnerability has been detected, but organizations are urged to apply the patch promptly to prevent unauthorized access. This disclosure follows previous findings by AppOmni on ServiceNow's AI platform vulnerabilities, highlighting ongoing security challenges in AI-driven environments. The incident serves as a reminder of the critical importance of timely patch management and proactive vulnerability assessments.

The UK Ministry of Defence has launched Project Nightfall to develop a ground-launched ballistic missile for Ukraine, aiming to counter Russian forces with a range exceeding 500 kilometers. The project seeks rapid production, awarding £9 million contracts to three industry teams to design and test missiles within 12 months, emphasizing speed due to Ukraine's urgent defense needs. Utilizing existing missile components, the project aims to create a UK version of the US Precision Strike Missile, enabling quick deployment and evasion of enemy retaliation. The missiles are designed for high-threat environments, capable of operating amidst heavy electromagnetic interference, with a production target of ten systems per month. The initiative seeks to minimize foreign export controls, likely to avoid political vetoes from nations sympathetic to Russia, ensuring uninterrupted support to Ukraine. Project Nightfall is expected to inform future UK military strategies, potentially integrating the developed technology into the British Army's long-range strike capabilities. The UK's artillery capacity has been diminished, partly due to support for Ukraine, highlighting the need for modernization and strategic enhancement of its defense systems.

Researchers at Securonix have uncovered SHADOW#REACTOR, a new malware campaign using a multi-stage attack chain to deploy the Remcos RAT, targeting enterprise and SMB environments. The attack begins with an obfuscated VBS launcher executed via wscript.exe, which initiates a PowerShell downloader to retrieve fragmented payloads from a remote host. Payload fragments are reconstructed in memory using a .NET Reactor-protected assembly, complicating detection and analysis efforts by security tools. The final stage employs MSBuild.exe as a living-off-the-land binary, enabling the Remcos RAT to establish covert control over compromised systems. The campaign's use of text-only stagers and PowerShell for in-memory reconstruction reflects a sophisticated approach to evade antivirus detection and sandbox analysis. SHADOW#REACTOR's tactics align with those of initial access brokers, indicating potential financial motivations, though no specific threat group has been identified. The campaign's self-healing design ensures resilience by allowing payload retrieval to continue even if initial attempts fail, enhancing its persistence capabilities.

CISA has identified active exploitation of a high-severity flaw in Gogs, adding it to its Known Exploited Vulnerabilities catalog. The vulnerability, CVE-2025-8110, involves path traversal in the repository file editor, with a CVSS score of 8.7. Attackers can exploit this flaw to execute code by manipulating symbolic links within the PutContents API. Approximately 700 Gogs instances have been compromised, with the majority of exposed servers located in China and the U.S. No official patch is available yet, but code changes have been made on GitHub, pending release in future updates. Users are advised to disable open-registration and restrict server access via VPNs or allow-lists as interim measures. Federal agencies must implement necessary mitigations by February 2, 2026, to protect against this vulnerability.