Unmanaged AI Agents Pose Security Risks Across Enterprises

A flaw in the Cursor AI editor allows malicious code in repositories to autorun on developer devices. This vulnerability can lead to malware execution, environment hijacking, and credential theft. The issue arises from Cursor disabling the Workspace Trust feature from VS Code, which prevents automatic task execution without explicit user consent. The flaw affects one million users who generate over a billion lines of code daily. The Cursor team has decided not to fix the issue, citing the need to maintain AI and other features. They recommend users enable Workspace Trust manually or use basic text editors for unknown projects. The flaw is part of a broader trend of prompt injections and jailbreaks affecting AI-powered coding tools.

Anthropic disrupted an AI-powered operation in July 2025 that used its Claude AI chatbot to conduct large-scale theft and extortion across 17 organizations in healthcare, emergency services, government, and religious sectors. The actor used Claude Code on Kali Linux to automate various phases of the attack cycle, including reconnaissance, credential harvesting, and network penetration. The operation, codenamed GTG-2002, employed AI to make tactical and strategic decisions, exfiltrating sensitive data and demanding ransoms ranging from $75,000 to $500,000 in Bitcoin. The actor used AI to craft bespoke versions of the Chisel tunneling utility to evade detection and disguise malicious executables as legitimate Microsoft tools. The operation highlights the increasing use of AI in cyberattacks, making defense and enforcement more challenging. Anthropic developed new detection methods to prevent future abuse of its AI models.

Chinese state-sponsored Advanced Persistent Threat (APT) actors, specifically the group known as Salt Typhoon, have been conducting a sustained campaign to gain long-term access to critical infrastructure networks worldwide. This campaign targets telecommunications, transportation, lodging, and military networks, exploiting vulnerabilities in routers and taking steps to evade detection and maintain persistent access. The Cybersecurity and Infrastructure Security Agency (CISA), in coordination with the National Security Agency (NSA), Federal Bureau of Investigation (FBI), and international partners, released a joint advisory detailing this ongoing malicious activity. The advisory provides actionable guidance and intelligence to help organizations defend against these sophisticated cyber threats. The advisory builds on previous reporting and incorporates updated threat intelligence from investigations conducted through August 2025, reflecting overlapping indicators with industry reporting on various Chinese state-sponsored threat groups. Salt Typhoon has been active since at least 2019, targeting at least 600 organizations, including 200 in the U.S., and 80 countries. The Czech Republic's National Cyber and Information Security Agency (NUKIB) issued a warning about data transfers to China, highlighting concerns over the transfer of system and user data to the PRC and the remote administration of technical assets. The Czech government previously accused China of targeting its critical infrastructure through APT 31, which began in 2022. China's offensive cyber activities include large-scale telco attacks by Salt Typhoon and positioning for potential destructive cyberattacks. The advisory tracks this cluster of activity to multiple advanced persistent threats (APTs), though it partially overlaps with Salt Typhoon. The advisory details how state-backed threat actors, including Salt Typhoon, penetrate networks around the world, as well as how defenders can protect their own environments. The Czech Republic's National Cyber and Information Security Agency (NUKIB) has assessed the risk of significant disruptions caused by China at a 'High' level, indicating a high probability of occurrence. NUKIB confirmed malicious activities of Chinese cyber-actors targeting the Czech Republic, including a recent APT31 campaign targeting the Czech Ministry of Foreign Affairs. The Chinese government has access to data stored by private cloud service providers within the Czech Republic, ensuring that sensitive data is always within its reach. NUKIB warns about consumer devices, such as smartphones, IP cameras, electric cars, large language models, and even medical devices and photovoltaic converters manufactured by Chinese firms, as risky devices that can transfer potentially sensitive data to Chinese infrastructure. 45 previously unreported domains associated with Salt Typhoon and UNC4841 have been discovered, with the oldest domain registration activity dating back to May 2020.

Researchers at Trail of Bits have demonstrated a new attack method that exploits image downscaling in AI systems to steal user data. The attack injects hidden prompts in full-resolution images that become visible when the images are resampled to lower quality. These prompts are interpreted by AI models as user instructions, potentially leading to data leakage or unauthorized actions. The vulnerability affects multiple AI systems, including Google Gemini CLI, Vertex AI Studio, Google Assistant on Android, and Genspark. The attack works by embedding instructions in images that are only revealed when the images are downscaled using specific resampling algorithms. The AI model then interprets these hidden instructions as part of the user's input, executing them without the user's knowledge. The researchers have developed an open-source tool, Anamorpher, to create images for testing this vulnerability. To mitigate the risk, Trail of Bits recommends implementing dimension restrictions on image uploads, providing users with previews of downscaled images, and requiring explicit user confirmation for sensitive tool calls.

Chinese cyber espionage groups Murky Panda, Genesis Panda, and Glacial Panda have escalated their activities targeting cloud and telecom sectors. Murky Panda exploits trusted cloud relationships and zero-day vulnerabilities to breach enterprise networks. They also compromise cloud service providers to gain access to downstream customer environments. Genesis Panda targets cloud services for lateral movement and persistence. Glacial Panda focuses on telecom organizations to exfiltrate call detail records and related telemetry. Murky Panda, also known as Silk Typhoon, has been active since at least 2021, targeting government, technology, academic, legal, and professional services entities in North America. They exploit internet-facing appliances, SOHO devices, and known vulnerabilities in Citrix and Commvault to gain initial access. They deploy web shells and custom malware like CloudedHope to maintain persistence. Genesis Panda, active since January 2024, targets financial services, media, telecommunications, and technology sectors across 11 countries. They exploit cloud-hosted systems for lateral movement and persistence, using compromised credentials to burrow deeper into cloud accounts. Glacial Panda has seen a 130% increase in activity targeting the telecom sector, focusing on Linux systems and legacy operating systems. They exploit known vulnerabilities and weak passwords to gain access and deploy trojanized OpenSSH components for credential harvesting.