Find what you’re looking for

ESET Threat Report H1 2026: Unveiling PromptSpy, the First AI-Driven Android Malware

In February 2026, ESET researchers discovered PromptSpy, the first known Android malware to utilize generative AI during its execution. This malware leverages Google's Gemini AI to interpret on-screen elements dynamically, enabling it to adapt its behavior across various Android devices and maintain persistence by preventing uninstallation. PromptSpy is distributed through a malicious dropper disguised as a system update, primarily targeting Spanish-speaking users in South America, especially Argentina. Once installed, it abuses Accessibility Services to monitor and control the user interface, deploys a Virtual Network Computing (VNC) module for remote access, and captures sensitive data such as lockscreen credentials and screen activity. ([eset.com](https://www.eset.com/us/about/newsroom/research/eset-research-discovers-promptspy-first-android-threat-using-genai/%3Fsrsltid%3DAfmBOoqZ_0fHGAaMnVaEZ5B0AuPdwhhXlaecY3Klyk-8QVRG-fAAlTy6?utm_source=openai)) The emergence of PromptSpy signifies a pivotal shift in mobile cybersecurity, illustrating how threat actors are integrating generative AI to enhance malware adaptability and persistence. This development underscores the urgent need for advanced detection mechanisms and proactive security measures to counteract AI-driven threats in the evolving cyber landscape.

AI's Breakthrough in Firmware Decryption: A Case Study

In 2026, Bishop Fox researchers utilized Anthropic's Claude, an advanced AI model, to autonomously reverse-engineer and decrypt SonicWall's proprietary firmware encryption. Without prior knowledge of the encryption format, Claude successfully traced the decryption logic, reconstructed the master key from embedded Shamir secret shares, and decrypted the firmware image. This achievement highlights the potential of AI in performing complex cybersecurity tasks traditionally requiring senior-level expertise. The experiment underscores the evolving role of AI in cybersecurity, demonstrating that AI models can independently execute sophisticated tasks such as firmware decryption. This advancement prompts a reevaluation of security strategies, emphasizing the need for continuous adaptation to AI capabilities in both offensive and defensive contexts.

Ubiquiti UniFi OS Vulnerabilities: Immediate Action Required

In July 2026, Ubiquiti disclosed seven critical vulnerabilities in its UniFi OS ecosystem, notably CVE-2026-50746, which allows network-based attackers to execute command injection attacks on devices managing smart building operations. These flaws affect multiple UniFi applications, including Connect, Talk, Access, and Protect, as well as various hardware devices. Exploitation could lead to unauthorized control over critical infrastructure components. The widespread exposure of over 100,000 UniFi OS instances online, particularly in the United States, underscores the urgency for immediate patching. Given the history of Ubiquiti devices being targeted to build botnets, these vulnerabilities present a significant risk to both individual organizations and broader network security.

CISA Mandates Immediate Patching of Critical Adobe ColdFusion Vulnerability CVE-2026-48282

In early July 2026, the U.S. Cybersecurity and Infrastructure Security Agency (CISA) mandated federal agencies to patch a critical vulnerability in Adobe ColdFusion, identified as CVE-2026-48282. This path traversal flaw affects versions 2025.9, 2023.20, and earlier, allowing unauthenticated remote attackers to execute arbitrary code on unpatched systems. Adobe released security updates on June 30, 2026, urging immediate deployment due to the high risk of exploitation. ([bleepingcomputer.com](https://www.bleepingcomputer.com/news/security/cisa-orders-feds-to-patch-max-severity-coldfusion-flaw-by-friday/?utm_source=openai)) The urgency of this directive underscores the rapid exploitation of such vulnerabilities by threat actors. Organizations must prioritize timely patching and robust vulnerability management to mitigate risks associated with critical software flaws.

CISA Urges Immediate Patching of Langflow Vulnerability CVE-2026-55255

In July 2026, the U.S. Cybersecurity and Infrastructure Security Agency (CISA) mandated federal agencies to patch a critical vulnerability in Langflow, a popular AI development tool. Identified as CVE-2026-55255, this Insecure Direct Object Reference (IDOR) flaw allows authenticated attackers to execute flows belonging to other users by manipulating the /api/v1/responses endpoint. Exploitation of this vulnerability can lead to unauthorized access to sensitive data and resource consumption. ([bleepingcomputer.com](https://www.bleepingcomputer.com/news/security/cisa-orders-feds-to-prioritize-patching-langflow-auth-bypass-flaw/?utm_source=openai)) The urgency of this directive underscores the increasing targeting of AI development platforms by cyber actors. As AI tools become integral to various sectors, ensuring their security is paramount to prevent potential data breaches and operational disruptions.

KDDI Data Breach 2026: Zero-Day Vulnerability Exploited

In June 2026, Japanese telecommunications giant KDDI detected unauthorized access to its email platform, affecting multiple internet service providers (ISPs) including STNet, JCOM, Chubu Telecommunications, NIFTY Corporation, and BIGLOBE. The breach, initiated on May 16, exploited a zero-day vulnerability in third-party software, leading to the exposure of approximately 12.23 million email addresses and 7.61 million passwords. KDDI promptly blocked the attackers upon discovery on June 17 and implemented defensive measures to secure the compromised systems. This incident underscores the critical importance of securing third-party software components, as vulnerabilities in such software can serve as entry points for attackers. Organizations are urged to conduct thorough security assessments of third-party tools and implement robust monitoring systems to detect and respond to unauthorized access promptly.

Entra Passkey Enrollment Vishing Targets Microsoft 365 Users

In April 2026, a threat actor identified as O-UNC-066, operating under the extortion brand 'Pink,' initiated a vishing campaign targeting Microsoft 365 users across multiple sectors, including food and beverage, technology, healthcare, automotive, construction, and aviation. The attackers impersonated IT personnel, contacting employees by phone and instructing them to enroll a new Microsoft Entra passkey for security purposes. Victims were directed to phishing websites mimicking legitimate Microsoft enrollment portals, where attackers captured credentials and multi-factor authentication (MFA) responses. Subsequently, the attackers registered passkeys under their control, gaining unauthorized access to victims' Microsoft accounts and exfiltrating data from services like SharePoint and OneDrive. ([bleepingcomputer.com](https://www.bleepingcomputer.com/news/security/entra-passkey-enrollment-vishing-targets-microsoft-365-users/amp/?utm_source=openai)) This incident underscores a growing trend of sophisticated social engineering attacks exploiting emerging authentication technologies. The use of real-time phishing kits capable of adapting to various MFA methods highlights the evolving tactics of cybercriminals. Organizations must remain vigilant, as such attacks can lead to significant data breaches and financial extortion. ([bleepingcomputer.com](https://www.bleepingcomputer.com/news/security/entra-passkey-enrollment-vishing-targets-microsoft-365-users/amp/?utm_source=openai))

Hackers Exploit Roundcube Flaw to Spy on Academic Researchers

In May 2026, a China-linked threat cluster, identified as UNK_MassTraction, exploited vulnerabilities in Roundcube webmail servers at U.S. and Canadian universities. Targeting physics and engineering departments, the attackers sent malicious emails that, when opened in vulnerable Roundcube clients, triggered the execution of JavaScript code exploiting CVE-2024-42009. This led to the deployment of IceCube malware, harvesting credentials and two-factor authentication data. Further exploitation of CVE-2025-49113 allowed the installation of SquareShell, a PHP webshell, granting remote code execution capabilities. In cases where this failed, the attackers deployed VShell, a Go-based backdoor facilitating interactive shell access and port forwarding. This incident underscores the persistent threat posed by state-sponsored cyber espionage, particularly targeting academic institutions involved in sensitive research areas. The exploitation of known vulnerabilities in widely used software like Roundcube highlights the critical need for timely patching and robust security measures to protect against sophisticated attacks.

Mount Royal University 2026 Ransomware Attack: A Case Study

In June 2026, Mount Royal University (MRU) in Calgary experienced a significant cyberattack attributed to the CMD Organization ransomware group. The attackers infiltrated MRU's network, exfiltrated data from the H drive—used by students and employees—and subsequently deleted the original files to hinder recovery efforts. This breach disrupted various university services, including online platforms and internal systems, affecting current and former students and staff. The university has engaged external cybersecurity experts and reported the incident to relevant authorities. ([bleepingcomputer.com](https://www.bleepingcomputer.com/news/security/mount-royal-university-confirms-breach-as-hackers-claim-attack/?utm_source=openai)) This incident underscores the evolving tactics of ransomware groups like CMD Organization, which employ auction-based extortion models to maximize financial gain. Their approach not only involves data encryption but also public data leaks and auctions, amplifying pressure on victims. ([labs.beazley.security](https://labs.beazley.security/articles/cmd-organization-new-ransomware-operator-moves-to-place-public-bidding-wars-on-ransomed-data?utm_source=openai))

Malicious SDKs on npm and PyPI Compromise Paysafe and Skrill Integrations

In July 2026, a coordinated supply-chain attack targeted developers integrating payment services by distributing at least 17 malicious packages on the npm and PyPI repositories. These packages masqueraded as legitimate SDKs for Paysafe, Skrill, and Neteller, aiming to steal sensitive credentials such as API keys, AWS keys, and GitHub tokens. The attackers employed typosquatting techniques, publishing packages with names closely resembling authentic ones, leading to unauthorized access and potential data breaches. This incident underscores the escalating threat of supply-chain attacks within open-source ecosystems. The attackers' ability to infiltrate multiple package managers simultaneously highlights the need for enhanced vigilance and security measures among developers and organizations to safeguard against such sophisticated threats.

Understanding and Mitigating System Prompt Leakage in AI Applications

In July 2026, AWS Security highlighted the persistent issue of system prompt leakage in generative AI applications. System prompts, which guide the behavior of large language models (LLMs), often contain sensitive information such as role definitions, behavioral guidelines, and API responses. Threat actors can exploit vulnerabilities to extract these prompts, potentially exposing proprietary data and compromising application integrity. Despite various mitigation strategies, complete remediation remains elusive due to inherent limitations in current AI systems. This underscores the need for continuous vigilance and adaptive security measures in AI deployments. The increasing prevalence of system prompt leakage incidents, as noted in the 2025 OWASP LLM Top 10, reflects a broader trend of sophisticated attacks targeting AI systems. Organizations must prioritize robust security frameworks to safeguard against evolving threats in the AI landscape.

Vidar Infostealer Exploits Malvertising to Target SMBs in 2026

In April 2026, a sophisticated malvertising campaign targeted consumers and small to midsize businesses (SMBs) globally by delivering the Vidar infostealer and XMRig cryptomining malware. Attackers lured victims with ads for cracked software, leading them to download password-protected archives that concealed a Go-based loader. This loader executed defense-evasion techniques, including an in-memory Antimalware Scan Interface (AMSI) bypass, before deploying Vidar to harvest browser credentials and XMRig to mine Monero cryptocurrency. The campaign utilized the Factory-v3 framework to generate unique binaries, complicating detection efforts. ([darkreading.com](https://www.darkreading.com/cyberattacks-data-breaches/vidar-infostealer-smb-malvertising-campaign?utm_source=openai)) This incident underscores the evolving tactics of financially motivated threat actors who combine multiple monetization strategies within a single infection. The use of malvertising, coupled with advanced evasion techniques, highlights the need for organizations to enhance their cybersecurity defenses against such multifaceted threats.