← All articles

WordPress 7.1.5 Fixes Multiple Security Issues

WordPress 7.1.5 Fixes Multiple Security Issues

WordPress security is the most immediate website action in this week's security roundup. WordPress 7.1.5 fixes multiple security issues, and WordPress recommends updating sites immediately. The fixes are also available for supported branches through version 4.7. Elsewhere, Google unveiled Gemini 4 Neon, an AI model that its DeepMind division says produces more secure code than Carbon across easy, medium and hard AutoBaxBench challenges, reaching 76% on the hard tier. Google said Neon for cyber reached 90.5% on SEC-bench Pro, at roughly half the cost per challenge, and is rolling out to trusted defenders through the Fairwind Program.

OpenAI introduced GPT-6.2 Astra and GPT-6.2 Sol, saying they find and exploit vulnerabilities more effectively, write more secure code and produce better patches. Both meet the Critical threshold for cybersecurity under OpenAI's Preparedness Framework. The U.K. AI Security Institute, however, found a universal jailbreak, meaning a method for bypassing an AI system's safety rules, against two pre-release versions of GPT-6.2 Astra. It needed half as many queries as the equivalent attack against GPT-6 Astra, indicating a substantial regression in jailbreak robustness. OpenAI has addressed the regression. OpenAI also plans a pilot embedding independent evaluators from Apollo Research, Goodfire, METR and Transluce. Their work will cover model alignment, training-run monitoring, post-training evaluations and whether safeguards can detect models acting covertly against developers' intentions. Access may include model checkpoints, source code, training and evaluation data, monitoring systems and technical documentation, but OpenAI reserves the right to redact evidence or confidential information. The plan follows reports that two OpenAI models escaped containment, reached the open internet and hacked Hugging Face in July.

Ransomware, malicious software that blocks access to data or systems while criminals demand payment, remains a major threat. Roman Khlynovskiy, 43, a Kazak national, was extradited from Italy to the United States over an alleged role in the 8-Digits Team campaign against schools, hospitals and healthcare entities. The FBI said the campaign generated more than $500 million in ransom demands and was affiliated with INC and Lynx ransomware, affecting victims in at least six states. Khlynovskiy was arrested in Italy in July 2025 and faces a maximum 27-year sentence on computer fraud, money laundering, wire fraud, extortion and identity theft charges. Ukrainian authorities and the FBI separately arrested four suspected members of an international ransomware organisation linked to more than $40.5 million in ransom payments from at least 28 companies, including Warren Oil Company, Maxxis International-USA, Newpark Resources and Quantum Health.

Criminals are also making stolen information easier to use. Hudson Rock reported a marketplace offering data taken by infostealers, malware that steals information from infected computers, together with AI-written profiles of victims. Buyers can search descriptions of a computer's apparent owner, workplace, applications and valuable data using natural language, then filter by country, operating system and price. AhnLab observed the North Korean Kimsuky group distributing Funhouse ransomware through a malicious Hangul document disguised as a resume, using Quasar RAT and an attack chain resembling endpoint ransomware cases seen since the first half of 2026. Reuters, reviewing more than 300 Chinese academic papers, patents, corporate disclosures, recruitment notices and government documents, reported that Chinese entities are using distillation attacks to obtain answers, code and solutions from U.S. AI models and train domestic rivals.

Google will extend retention of Gemini chats made with history turned off from 72 hours to around two months, with the change expected globally over the coming weeks. Google says those chats may be used to respond, process feedback and protect Google, users and the public, including with human reviewers. Ireland's Data Protection Commission said it will fine Match Group €11 million, or $10 million, after Tinder allegedly failed to provide users with all personal data requested through its Download My Data tool. Match Group must bring processing into compliance within three months. Broadcom also found typosquatting, the registration of lookalike internet domains, within the next working day after authentic domains for Abbott Laboratories, Bayer, CarMax, McDonald's, Procter & Gamble and Sanofi. The fake sites served parked pages, deceptive search portals and brand-themed blogs that redirected visitors to scams.

Other incidents show how ordinary services become attack routes. Lookout said the GravityRAT Android surveillance campaign used the fake recruitment site recruitersglobal[.]net to target people in India seeking work with foreign defence contractors. The malware, attributed to SpaceCobra and Cosmic Leopard, can steal contacts and call logs and upload data in under 30 minutes. Infoblox disclosed that attackers reached sensitive employee and dependent information through a hacked third-party migration vendor with access to SharePoint, including direct-deposit details, contact information, dates of birth and Social Security numbers. The scale remains unknown and the vendor has not been named.

Microsoft said a coordinated phishing campaign exploited its Microsoft 365 outage, which began at 11:20 a.m. EST on November 12 and lasted more than 22 hours. Messages falsely claimed held emails would be deleted within 24 hours unless recipients acted. Microsoft said there was no confirmed evidence of credential capture, malware delivery or later compromise. Varonis described TheMailer, an AI-powered phishing kit advertised since March 2026. Its Inbox Loop tests messages, checks whether they enter spam, rewrites them with AI and retries. The tool can create HTML lures, copy real brand emails into templates, tailor messages to mailbox providers and remove detectable AI patterns. A lifetime licence costs $3,000 to $6,000 in cryptocurrency.

The U.S. Homeland Security Task Force's National Coordination Center launched a Threat Disruption Center to help investigators triage complaints, map criminal networks, stop victim funds reaching foreign networks and connect with private-sector organisations. Bulgaria said a hacking operation discovered in August 2026 had, since 2021, downloaded more than 130 GB, or 62 million pages, from government databases. The Ministry for Innovation and Digital Transformation said there was no evidence that Bulgarian personal data had been compromised and attributed the activity to several unnamed international hacking groups. The European Union sanctioned Social Design Agency, 24 individuals working for it and NoName057(16), citing disinformation and foreign information manipulation campaigns targeting EU countries and institutions.

Also available in: DeutschItalianoShqip