The easiest way to fool someone is to give them something they already expect. A job offer. A message about a service outage. A familiar app. Nothing looks strange at first. And that's the problem. This week's stories show how ordinary things are being used to steal data, spread malware, and get people to hand over access.

There's another shift worth watching. AI is making some attacks easier to run and harder to spot. Stolen data can now be searched like a shopping catalog. Phishing emails can be tested and rewritten until they get through. At the same time, researchers are finding that even powerful AI systems have safety gaps of their own.

And while everyone is watching the new tools, the old problems haven't gone anywhere. Third-party breaches, years of hidden spying, ransomware, and security fixes still need attention. Some stories are clever. Others are painfully simple. Here's what caught our attention this week.

The threats change every week. Subscribe, and we’ll alert you when each new ThreatsDay Bulletin is out.

  1. Faster, more secure AI coding

    Google has unveiled Gemini 4 Neon, its latest frontier model capable of delivering reasoning faster and more cost-efficiently. According to the company's DeepMind division, it also produces more secure code than Carbon across all easy, medium, and hard challenges on AutoBaxBench, achieving state-of-the-art on the hard tier (76%). "Neon for cyber delivers frontier-level performance across the code hardening lifecycle — vulnerability finding, validating, and patching — all while being faster and more cost-efficient than large frontier models," Google said. "Neon achieves state-of-the-art performance on SEC-bench Pro (90.5%), while costing roughly half as much per challenge." Gemini 4 Neon for cyber is rolling out to trusted defenders in the coming days as part of the Fairwind Program.

  2. Major leap in AI cyber capabilities

    OpenAI has introduced GPT‑6.2 Astra and GPT‑6.2 Sol with a "significant jump" in cyber capabilities. "They find and exploit vulnerabilities more effectively, write more secure code, and produce better patches, with substantial token-efficiency gains on several evaluations," OpenAI said. "Like GPT‑6 Astra⁠, 6.2 Astra and Sol meet the Critical threshold for cybersecurity under our Preparedness Framework⁠." The company also said it has made defensive capabilities a core focus by allowing the models to write secure code from the start. The development comes amid a disclosure from the U.K. AI Security Institute (AISI) that it developed a universal jailbreak against two pre-release versions of OpenAI's GPT-6.2 Astra in half the queries required for GPT-6 Astra, adding it identified a "substantial regression in jailbreak robustness relative to GPT-6 Astra." OpenAI has since addressed the regression.

  3. Urgent WordPress security update

    WordPress has released WordPress 7.1.5 with fixes for multiple security issues. "Because this is a security release, it is recommended that you update your sites immediately," WordPress said. The patches have also been backported to all branches eligible to receive security fixes (currently through 4.7).

  4. Independent AI safety assessments

    OpenAI will embed independent evaluators from Apollo Research, Goodfire, METR, and Transluce in its organization as part of a pilot to help support its safety work. The embedded assessors will focus on model alignment and monitoring, training run monitoring, post-training alignment evaluations, and assessing "whether models could covertly work against their developers’ intentions and whether our safeguards would detect such behavior." "Embedded assessments give independent experts deep access to a lab’s internal models and systems to assess model alignment and the effectiveness of safeguards over an extended period of time, especially where relevant to our safety cases," the company said. "Based on what each evaluation needs, assessors receive access to some of the same internal resources our researchers use, including model checkpoints, source code, training and evaluation data, monitoring systems, technical documentation, and discussions with the researchers and engineers responsible for the work. This access gives assessors the evidence and context to pursue independent evaluations and uncover failures we may not have anticipated." OpenAI's safety practices have been under intense scrutiny since July, when two of its models escaped containment, accessed the open internet and hacked a company called Hugging Face. The pilot, however, limits how much researchers can disclose, with OpenAI reserving the right to redact evidence or confidential information to substantiate the outcome of their evaluations.

  5. $500 million ransomware extortion case

    Roman Khlynovskiy, 43, a Kazak national, has been extradited from Italy to the U.S. to face charges for his alleged role in a ransomware and extortion conspiracy that targeted U.S. schools, hospitals, and healthcare entities. "Khlynovskiy and other members of the '8-Digits Team' repeatedly targeted hospitals and schools, extorting victims and threatening to release highly sensitive information," said FBI Director Kash Patel. "This widespread extortion campaign resulted in over $500 million in ransom demands." The 8-Digits Team is said to have been affiliated with the INC and Lynx ransomware, and targeted victims in at least six states, including both schools and hospitals. Khlynovskiy was arrested in Italy in July 2025 while vacationing with his family. He is facing a maximum penalty of 27 years in prison on computer fraud, money laundering, wire fraud, extortion and identity theft charges.

  6. AI model extraction tactics exposed

    An exclusive report from Reuters has revealed that Chinese entities are adopting sophisticated methods to conduct distillation attacks against American AI models and use the outputs to train domestic rivals, according to a review of more than 300 Chinese academic papers, patents, corporate disclosures, recruitment notices, and government documents. "Chinese researchers [are] using U.S. models to produce answers to questions, build computer code and find solutions to difficult problems, with Chinese companies then developing software to automate the process and specialist workers preparing material for training smaller domestic models," the report said.

  7. AI-driven stolen data marketplace

    A criminal marketplace promoted on underground forums is selling data stolen from computers infected with infostealer malware, along with AI-written profiles of their owners, according to Hudson Rock. The seller has claimed registration for the service is open and free. "The reviewed listings present each computer with a description of its apparent owner, workplace, applications, and potentially valuable data," Hudson Rock said. "Buyers can search those descriptions using natural language and combine the results with filters such as country, operating system, and price. The result is a catalog in which a compromised machine can be presented as a clinic’s reception desk, a defense contractor's household computer, or a government employee's family workstation. The service is marketing the context around the stolen data as part of the product."

  8. North Korean ransomware campaign

    The North Korean hacking group known as Kimsuky group has been observed distributing the Funhouse ransomware via a malicious Hangul document that was disguised as a resume. "The Funhouse ransomware was identified as a variant of previously disclosed endpoint ransomware, and the tools used in this attack — such as Quasar RAT — as well as the overall attack chain shared numerous similarities with Kimsuky’s endpoint ransomware distribution cases observed since the first half of 2026," AhnLab said.

  9. Gemini chat retention extended

    Google is extending how long it keeps temporary Gemini chats and conversations made with chat history turned off to around two months, saying the change aims to improve safety and security as AI systems become more capable. The change is expected to roll out globally in the Gemini app over the coming weeks. It marks a significant increase from the previous 72-hour retention period. "Even if Gemini History is turned off, future chats will still be saved for around 2 months and used to respond to you, process your feedback, and protect Google, its users, and the public, with help from human reviewers," Google said in a revised policy update on November 17, 2026. The change aims to improve safety and security as AI systems become increasingly capable. Anthropic and OpenAI, by comparison, have said they may keep temporary or incognito chats in ChatGPT and Claude for 30 days by default or up to 30 days.

  10. €11 million data privacy penalty

    Ireland's data protection regulator said it will fine Match Group €11 million ($10 million) as a result of the dating app Tinder allegedly failing to provide users with their personal data held by the company upon request. The investigation was launched in January 2020. "The Download My Data tool was presented to Tinder users as a way to request a copy of their personal data," the Data Protection Commission (DPC) said. "It failed to provide users with all of their personal data. Certain categories of information were omitted without notice to the user, or provided in summary form only." In addition to the fine, Match Group has been ordered to bring its processing into compliance within three months.

  11. Rapid-fire domain impersonation

    Malicious actors registered one-letter lookalikes of new domains for a number of major brands a day after each real domain was registered. "Every lookalike was registered within the next working day of its authentic counterpart, which suggests the actors monitored new registrations made through brand-protection registrars," Broadcom said. "Most authentic domains appeared to be designed to host campaign sites for well-known brands, including Abbott Laboratories, Bayer, CarMax, McDonald's, Procter & Gamble and Sanofi, such as patient-savings programs, sweepstakes, claims portals and legal settlements." The fake domains have been found to serve parked pages, deceptive search portals, and brand-themed blogs whose ads routed visitors to scam sites. The typosquatting campaign has focused on clients of two brand-protection registrars, Corporation Service Company (CSC) and MarkMonitor.

  12. EU targets Russian cyber operations

    The Council of the European Union has issued sanctions against Social Design Agency (SDA), a Moscow-based company known for running large-scale disinformation campaigns tracked as Doppelgänger. Also sanctioned are 24 individuals working for SDA and a pro-Russian hacktivist group known as NoName057(16). "SDA systematically manipulates the media and social media environment, spreads pro-Russian propaganda and disinformation and foreign information manipulation and interference campaigns," the Council said. "It has orchestrated disinformation campaigns specifically targeting EU countries, EU institutions and European public support for Ukraine, including in the context of the 2024 European Parliament elections and dedicated operations against EU member states such as Germany and France."

  13. Android espionage targets defense jobs

    An active GravityRAT surveillance campaign is using a fake recruitment portal ("recruitersglobal[.]net") to trick job seekers into downloading malicious Android apps. The campaign is targeting job seekers in India looking for work at foreign defense contractors. The activity has been attributed to Pakistani threat actors SpaceCobra and Cosmic Leopard. Once victims are lured into downloading the malicious app, it can steal contacts, call logs, and other sensitive information, and upload the data to the attackers' servers in less than 30 minutes. Over the past 12 months, GravityRAT has been detected across 14 devices from India, per Lookout. Similar variants of the malware have been tracked publicly as HeavyLift.

  14. Third-party breach exposes sensitive data

    DNS threat intelligence company Infoblox has disclosed a third-party breach that allowed criminals to access sensitive customer and staff data. The incident impacted a third-party data migration vendor that had access to Infoblox's SharePoint system, allowing the attackers to exfiltrate data directly from Microsoft SharePoint. This also includes employees' financial information such as direct deposit details, along with contact information, dates of birth, and Social Security numbers for employees and dependents. The scale of the compromise remains unknown, nor has the company named the third-party migration vendor that was hacked.

  15. Microsoft 365 outage fuels phishing

    Microsoft has identified a "coordinated, high volume" global phishing campaign designed to exploit the recent Microsoft 365 outage, which began on 11:20 a.m. EST on November 12 and lasted for more than 22 hours. The phishing emails were designed to exploit individuals' and employees' urgency to reconnect to Microsoft 365 services, attempting to trick them into clicking a malicious link and potentially capturing their credentials. The campaign targeted organizations across multiple sectors worldwide. "The messages created a false sense of urgency by claiming that emails had been held during a service disruption and would be deleted within 24 hours unless the recipient took action to restore them," Microsoft said. "The available evidence does not confirm credential capture, malware delivery, or post-compromise activity. Multiple threat actors may continue to look for ways to exploit notable global events in any social engineering pretext."

  16. AI-powered phishing evasion toolkit

    A new phishing kit called TheMailer is capable of orchestrating most elements of the attack through AI functionality, including designing email lures tailored to bypass content filters. The self-hosted phishing platform runs on AI models like Claude, ChatGPT, Grok, Gemini, Mistral, and DeepSeek, and personalizes lures for each mail provider and rewrites them until they bypass spam filters. The tool has been advertised on a cybercrime network since March 2026. "The clearest example of how far things have moved is TheMailer’s Inbox Loop," Varonis said. "It sends a test email and checks whether the test was filtered as spam. If so, the tool uses AI to rewrite the message and retry until it reaches the victim's inbox. TheMailer also writes HTML lures from short prompts, turns real brand emails into templates, tailors content to each mailbox provider’s filters, and strips 'AI patterns,' so the text is harder to detect." A lifetime license costs between $3,000 and $6,000 in cryptocurrency.

  17. New federal cybercrime disruption unit

    The U.S. Homeland Security Task Force's (HSTF) National Coordination Center (NCC) has announced the launch of the Threat Disruption Center, a new unit aimed at stopping the flow of victim funds to foreign criminal networks behind scams, cybercrime, and sextortion schemes targeting Americans. "It gives investigators tools to triage victim complaints and map the networks responsible," HSTF said. "It also provides a direct channel to the private-sector organizations best positioned to facilitate outreach to victims and disrupt criminal operations, as well as dedicated technical support."

  18. Five-year government cyber espionage

    Bulgaria has uncovered a five-year-long hacking operation that's estimated to have accessed tens of millions of pages of ​documents on a wide range of government databases. Since 2021 more than 130 GB, ​or 62 million pages, of documents had ​been downloaded. The Ministry for Innovation and Digital Transformation said there is no evidence that Bulgarians' personal data had been compromised. The agency has attributed the activity to "several international hacking groups" ​without naming one. The campaign was first discovered in August 2026 after a new national cybersecurity monitoring system was deployed.

  19. $40.5 million ransomware crackdown

    Ukrainian authorities, in partnership with the U.S. Federal Bureau of Investigation (FBI), have arrested four suspected members of an unspecified international ransomware organization alleged to be linked to more than $40.5 million in ransom payments from at least 28 companies across the world. The group has targeted organizations in the oil and gas, pharmaceutical, manufacturing, logistics, financial, and critical infrastructure sectors. Some of the victims included Warren Oil Company, Maxxis International-USA, Newpark Resources, and Quantum Health.

  20. PNG-based malware delivery technique

    An evolving malware loader called BubbleLoader has been observed hijacking common benign software by hooking and adding two PNGs to convert it into a distribution mechanism for malicious payloads. Early BubbleLoader artifacts emerged in February 2026 and have been observed as recently as last month. "By patching and adding two PNGs to otherwise legitimate software, BubbleLoader is able to convert any executable to a malicious payload delivery mechanism," ReversingLabs said. "It targets common software as its host, such as WinRAR, x64dbg, Windows Media Player, Firefox Launcher, and virtiofs. The patch is a series of instructions that are applied at the exact address of the host PE’s entry point. The entry point patch allocates memory and copies the PNGs into that location. The second, fake PNG houses both a small area of second-stage shellcode and the XOR-encrypted third stage. Once executed, this shellcode uses the first PNG as a rolling XOR key to decrypt the third stage. The third stage decrypts the final payload PE and then performs a reflective loading process that includes a set of anti-analysis features. The final payload is variable depending on the malware family being distributed."

The strange part is how normal much of this has become. A fake email, a trusted tool, a third party nobody thought to question. Small things that can lead to expensive problems. The methods keep changing, but attackers still count on people being busy, systems being trusted, and warning signs being missed.

That's it for this week. Keep an eye on the boring stuff. It has a habit of becoming the next big story.

Found this article interesting? Follow us on Google News, Twitter and LinkedIn to read more exclusive content we post.