AI, IoT, and Critical Infrastructure Under Fire: Daily Cybersecurity Brief

The ongoing cyber campaign by Iran-linked actors targeting US water and wastewater facilities has expanded its scope. Initially reported in at least seven states, the attacks now affect at least a dozen states, with New Jersey and Alabama explicitly named as new victims. Beyond operational disruptions like pressure loss and flooding, new reports indicate disruptions to phone and communication systems. Technical details now highlight specific target ports 44818 (EtherNet/IP) and 2222 (Telnet/CLI) for Rockwell devices, emphasizing the exploitation of internet-exposed PLCs via default credentials and remotely accessible services.

The estimated financial loss from the Coldcard wallet vulnerability has increased significantly, now ranging from $100 million to $130 million. Further technical analysis reveals the flaw, a build configuration error, critically reduced the seed phrase entropy from 128 bits to just 40 bits, making private key brute-forcing feasible. Exploitation is now attributed to multiple distinct hacking groups, indicating a broader attack surface and impact, including high-profile victims. This update highlights the escalating severity and wider scope of the incident.

Further analysis of the 'RovoBlast' vulnerability in Atlassian Rovo reveals the specific 'Parameter-to-Prompt (P2P)' injection technique used to bypass traditional jailbreaking. Attackers could leverage markdown image rendering for data exfiltration. The scope of potentially affected platforms is clarified to include Microsoft 365 and Google Workspace, in addition to Jira and Confluence. Importantly, despite the critical nature of the flaw, there is currently no evidence of this vulnerability being exploited in the wild, reinforcing that Atlassian's patch was timely and effective. Organizations are reminded to apply least privilege principles to AI agents.

Unit 42 reports Kimwolf v7, an evolution of the AISURU botnet family, now targets Android-based IoT devices via ADB exploitation. This variant features a sophisticated HTTP/2-based DDoS flood that mimics legitimate browser traffic, making detection and mitigation more challenging. Its command and control (C2) infrastructure has been significantly enhanced for resilience, utilizing the Ethereum Name Service (ENS) for primary C2 resolution and a Tor .onion address as a fallback. These advancements represent a direct response to previous C2 takedown efforts, increasing the botnet's operational durability and overall threat level.

The ongoing TeamPCP campaign has escalated with a significant attack on the LiteLLM open-source AI framework. Malicious versions 1.82.7 and 1.82.8 were published to PyPI, designed to harvest cloud credentials, API keys for LLMs (OpenAI, Azure, Google), and cloud providers (AWS, GCP) from AI development environments. This incident, leveraging .pth files for execution, is considered the largest AI infrastructure breach of 2026, affecting over 2,500 organizations and 434,000 CI/CD pipelines. The FBI warns of long-term credential misuse.

New information clarifies that the October 2025 incident at Unlimited Systems was a ransomware attack where an unauthorized actor accessed and exfiltrated sensitive patient data before deploying ransomware. This confirms a double extortion tactic, adding a critical technical detail to the initial report of unauthorized access. Furthermore, the significant nine-month delay in notifying the 3.8 million affected individuals has now resulted in a proposed class-action lawsuit against the company. This legal action significantly increases the incident's impact and consequences for Unlimited Systems, highlighting the severe repercussions of delayed breach disclosures and the nature of the attack.

Further analysis of Storm-1175's campaign exploiting CVE-2026-18577 reveals additional post-compromise tactics. The threat actor is now observed using Advanced IP Scanner for network discovery and Mimikatz for credential dumping from LSASS process memory. This indicates a more sophisticated approach to lateral movement and persistence within compromised networks. N-able has also released a second hotfix for CVE-2026-18577, emphasizing the critical need for immediate patching to address the authentication bypass vulnerability.