The weekly SecurityWeek overview of cybersecurity news provides a concise overview of important developments that may not receive comprehensive, stand-alone coverage, but remain relevant in the broader threat landscape.
This curated summary highlights top stories regarding vulnerability disclosures, emerging attack methods, policy updates, industry reports, and other noteworthy events to help readers stay well-informed about the evolving cybersecurity environment.
Here are this week’s highlights:
Iran suspected of involvement in hacking US military freezers
U.S. officials suspect hackers linked to Iran caused freezer malfunctions that spoiled food at about six military base commissaries starting in late August, Nextgov reports, citing people familiar with the matter. At Fort Huachuca, Arizona, a freezer went into defrost mode and heated its contents, and law enforcement from Cyber Command and the Department of Defense are investigating with French authorities, after hackers overrode refrigeration controls at some 350 French facilities earlier this month.
Most servers rely on UEFI code affected by critical boot flaw
A 16-year-old critical buffer overflow (CVE-2026-97631) in TianoCore EDK II, open source code that most server vendors and data centers around the world use through firmware vendors such as AMI, Insyde and Phoenix, means the flaw affects most servers around the world, Ars Technica reports. An attacker with access to the local network can send an oversized ICMP error message to a machine while it is booting over the network (PXE) to suspend the process or, if the target’s firmware version is known, execute code at the highest privilege level before defenses such as secure boot take effect. Only machines using network boot are exposed, and the flaw, discovered by researchers at Anthropic and Google, became public months before its planned disclosure in January after AMI released a code patch, forcing organizations to seek firmware updates from their vendors.
Anthropic challenges scope of ex-startup DOGE’s AI-based intrusion
Cathedral, a cybersecurity startup founded by former DOGE members, says its AI agents found a way into Anthropic’s internal systems in less than 12 hours by abusing the weaknesses of third-party enterprise software, accessing meeting transcripts, salary and job posting data, and Slack chats on the company’s most advanced models, according to a draft report reviewed by the Washington Post (paywall). Anthropic, which received the results through its bug bounty program, claims that only a limited set of business, HR and IT systems were hit, that it cut off the startup’s access the same day and that Claude’s conversations, source code and model weighting were not affected. It also accuses Cathedral of violating the rules of the bounty program by not disclosing all of its activities.
Midterm spending disclosures stall after FEC intrusion attempt
The US Federal Election Commission took its electronic filing system offline after someone attempted to breach its systemsaccording to an internal FEC email seen by Reuters, which prevents political groups from submitting campaign finance reports a week before the Nov. 3 midterm elections. The outage blocked 24-hour and 48-hour reports that track last-minute spending on specific races, although past records remain searchable. The system was restored on October 28.
Hackers Obtain National ID Numbers, Card PINs in University of Copenhagen Breach
THE University of Copenhagen claims that hackers who gained access to one of its testing environments obtained personal data on approximately 525,000 current and former members of the university community, including names, phone numbers and personal email addresses, as well as national identity numbers (CPR) for many of them. The attack was detected and contained on October 20, but the ID card PINs of around 20,000 staff and students were also compromised and have since been reissued. The university says it does not yet know who was behind the attack or what the motive was.
ENGIE data breach attributed to a tool used by business partners
French energy giant ENGIE declares attackers stole customer data after having had access to a tool that its partners use for customer prospecting and the marketing of energy offers, but that its own internal IT systems have not been compromised. The incident is limited to customers in France, and notification letters indicate that the stolen data includes names, contact details, customer IDs, contract numbers and energy supply point IDs, but no banking information. ENGIE did not specify how many customers are affected, and no group has claimed responsibility for the attack.
Fake bots ChatGPT and Claude probe sites for leaked secrets
Cequence tracked a single threat actor who alternates with approximately 32 AI bot identities (including crawlers and helpers from OpenAI, Anthropic, Meta, and Perplexity) while scanning websites for files that typically contain secrets, such as .env files, cloud credentials, and private SSH keys. Running by a major cloud provider since at least the end of September, the scanner used more than 300 IP addresses in a single day and, in one day in October, reached 48% of organizations protected by Cequence that saw AI bot traffic. Sequence warns that a bot’s self-reported identity is unreliable, noting that the impersonator accounted for more than 80% of traffic labeled as AI bot traffic in some organizations.
Cloudflare and SGLang fix cross-tenant data leaks in AI inference
Cloudflare says that between August 27 and September 9, 155 responses from the GLM 5.3 model on its Workers AI platform contained data from recent requests from other clients to the same pattern, all involving prompts over 681,280 tokens. The leak resulted from a configuration mismatch that left the model’s KV cache smaller than its advertised context window, combined with a bug in the SGLang inference engine that allowed overly long queries to read cache data left over from previous sessions. Cloudflare fixed the issue after Knostic researcher Gadi Evron reported it via HackerOne, found no evidence of deliberate exploitation, and recommended SGLang operators update to version 0.5.22 or later.
Russian defense employee and investors targeted by North Korean IT project
John Allen Kurrle Jr., 62, an employee of a defense contractor with Top Secret clearance, pleaded guilty for wire fraud conspiracy to assist a North Korean IT telecommuter program by allowing foreign individuals to use his identity (and that of a third party) to apply for remote jobs and by hosting victim companies’ laptops in his home. The applications resulted in nine job offers from U.S. companies and Kurrle faces up to 20 years in prison. Furthermore, OFAC sanctioned Ten individuals and one entity support North Korea’s IT worker projects, including two Russians who allegedly converted workers’ crypto earnings into cash, two North Koreans who managed Russian-based IT worker groups, and the Russia-based First Financial and Technical Development Center (FFTDC) which deploys North Korean IT workers overseas.
Rockwell targets SecureOT at OEMs facing CRA and machinery rules
Rockwell Automation added OEM-Focused Capabilities to its SecureOT offering to help machine builders prepare for cybersecurity requirements under the Cyber Resilience Act (CRA) and EU Machinery Regulations. Built on the SecureOT platform, additions include machine-level hardware, software and firmware inventories, vulnerability identification, continuous monitoring of newly revealed vulnerabilities after deployment and exportable compliance reports for customers and auditors. Rockwell also offers assessment and consulting services to help OEMs identify gaps, establish governance processes and align with standards such as IEC 62443.
AI Tackles Ransomware, C2 Malware, and Gray Market Tools
Sophos claims Gold Ogre Group is behind Aura0ra Ransomware made extensive use of AI, with agents orchestrated through Cursor handling network discovery, credential attacks, and lateral movement with limited human involvement, after operators circumvented guardrails by presenting intrusions as authorized red teamwork. Cisco Talos detailed NIK C2a new AI-integrated command and control framework whose Windows implant can steal credentials, broadcast keystrokes, and deploy screen lock ransomware. Symantec and Carbon Black, for their part, warn of a gray market relay service, seen in a dozen organizations primarily in mainland China, which installs the real Codex and Claude Code CLIs but reconfigures them to route all requests through third-party servers with sandboxing disabled.
Related: In Other News: ICS-Sniper Attack, Hack of XWorm Malware Operators, $250,000 Chrome Exploit
Related: In Other news: Tor output relays hijacked, banks spot malicious AI agents, self-rebuilding macOS ransomware
