Week in review: Gyazo breach exposes 23.6M user data, TASK#STOMP steals documents

Here’s an overview of some of last week’s most interesting news, articles, interviews and videos:

Know what was tested before your SAP ECC migration goes live
In this Help Net Security interview, Guilherme Joventino, COO of MIGNOW, explains why some large companies plan to stay on ECC past the 2027 deadline and pay SAP for extended support until 2030. The interview covers what that choice may cost, why fear of disruption stalls projects more often than budget, and what the first ninety days of a phased migration involve.

The TASK#STOMP Windows backdoor takes Wi-Fi passwords, screenshots, and business files
Researchers have taken apart TASK#STOMP, a Windows backdoor that searches a victim’s drives for business documents, uploads them to attacker servers, and then stays put to grab each new or edited document. The same malware steals saved Wi-Fi passwords and clipboard text, takes screenshots, and runs whatever command its operators send.

European AI spending is on track to reach nearly $470 billion by 2030
European organizations will spend nearly $470 billion on AI in 2030, IDC forecasts, with spending growing at a compound annual rate of 35% from 2025. At that rate, the market more than quadruples in five years. Generative AI will account for 55.4% of the total by 2030.

Somewhere in your traffic logs, a bot is doing more than looking
Akamai has watched verified AI crawlers, ChatGPT among them, move from reading web pages to sending high-frequency POST requests. In a 30-day analysis of its global customers, ecommerce accounted for 44.8% of those AI bot POST transactions, and travel climbed to 30% in a single month.

What to do first when you get 90 days to secure AI agent data
In this interview with Help Net Security, Kelly Herrell, CEO at Nol8, explains where AI agents create exposure inside organizations. The first thing to examine is the data path: what an agent can reach, what enters its context, and where results go.

Stop watching what AI agents say and start watching what they do
In this interview with Help Net Security, Ariel Assaraf, CEO of Coralogix, explains why a system prompt can describe a boundary for an AI agent but cannot enforce one. Assaraf covers how his team builds AI agent guardrails at the execution layer, limits context without expanding authority, and tunes controls by risk so agents stay useful during incidents.

North Korea’s job interview scam runs both ways
Attackers are targeting members of the Rust Project and maintainers of widely used crates (Rust code libraries), dangling attractive opportunities to compromise their devices and accounts and, ultimately, publish malware.

Attacker compromised nearly 1000 Zyxel switches since August (CVE-2026-7273)
A Chinese-speaking threat actor has exploited a vulnerability (CVE-2026-7273) in unpatched ZyXEL GS1900 Smart Managed Switches and has exfiltrated sensitive data from 996 devices across 48 countries, GreyNoise reported on Monday. The affected switches are predominantly located in Italy, the US, Taiwan, South Korea, and a number of EU countries.

Brief hijack makes Elsevier domains redirect to LAPSUS$ “Chapter II” page
Three domains / web portals belonging to Dutch academic publishing company Elsevier have been redirecting users to a page branded “LAPSUS$ GROUP, Chapter II,” carrying a signed statement that taunted the FBI and counted down to a future victim. According to Cloudskope researchers, the redirect ran for at least 78 minutes, from roughly 7:49pm CT until it was cleared before 10:09pm CT on September 21, 2026.

Attackers hit Check Point Management Servers and Spark firewalls, F5 BIG-IP APM instances
Check Point Software has released emergency fixes for a critical Check Point Management Server vulnerability (CVE-2026-93616) that has been exploited as far back as July 23, 2026. The company also confirmed that a pre-authentication remote code execution (RCE) vulnerability (CVE-2026-85102) in Check Point (Quantum) Security Gateway for which it released patches on September 9, 2026, started getting probed a few days after.

DarkMe RAT trades zero-days for plain phishing emails
DarkMe, a remote access trojan and info-stealer that has previously been associated with a threat group that targeted financial market traders and cryptocurrency users, has been spotted again. This time around, its distribution has been simplified: instead of leveraging zero-day exploits, attackers are betting on a simple email to convince targets to run it on their machine.

OpenAI agent hacking spree widens to Australia, targeting government website
Before the Hugging Face and RubyGems hacks, autonomous OpenAI agents attempted to hack into three other websites, including an Australian government public health website, independent research lab Transluce revealed on Wednesday. Insight into the agents’ actions was gleaned from reports of tens of thousands of queries apparently made by the agents through urlquery.net, a free URL scanning service, so they could avoid access restrictions.

Passwork NIS2 efficiency guide: Save your team hours before the 2026 audit
EU authorities are reviewing NIS2 compliance, increasing accountability for senior management while cybersecurity teams face skills shortages. This guide explains how Passwork supports NIS2 Article 21 requirements, reduces operational workload, and simplifies audit evidence collection.

Intent injection attacks are a new worry for AI-native 6G networks
Intent-based networking (IBN) lets operators state the outcome they want and leaves its translation into network policy to software, an approach AI-native 6G designs have moved to the forefront. Researchers at the University of Ottawa and Nokia Bell Labs argue that this abstraction gives attackers new openings, and it tests two machine-learning detectors against one of them.

Hackers exploit Gyazo server flaw to steal 23.6 million user records
Japanese software company Helpfeel has confirmed a data breach on its screenshot-sharing platform Gyazo, in which attackers exploited a vulnerability in its image upload server, stealing approximately 23.62 million user records and metadata tied to hundreds of millions of images.

Scammers impersonate cops, use arrest threats to extort victims
Scammers are posing as police officers and federal agents, threatening arrest unless victims pay up, the FBI warns. The FBI’s Internet Crime Complaint Center (IC3) updated an alert it first issued in 2022, citing “losses totaling more than $1.6 billion” between January 2025 and July 2026.

Scammers use genuine Google sign-ins to sell costly, unverified AI subscriptions
Scammers are using a $249 website toolkit to sell unverified AI subscriptions worth up to $2,000 a year, and a genuine Google sign-in screen is what makes the sites convincing. Malwarebytes found more than 100 websites built this way, all tied to the same toolkit and closely related developer details.

The latest deepfake numbers give CISOs plenty to worry about
AI is letting cybercriminals reach deeper into organizations than a phishing email ever could. 41% of CISOs reported at least one social engineering incident involving a deepfake during an employee audio call in the past 12 months, according to Gartner. 36% reported the same for a video call.

Researchers uncover malware that uses AI to choose its next move
To help security practitioners catch malware that leans on AI, researchers from Cisco Talos shared an open-source framework that they hope will be used to classify and analyze the threat. The tool, called CAIRN, works entirely from metadata pulled off files. No downloading the malware, no running it.

Microsoft disrupts EvilTokens phishing service that gave criminals access to 12,000 inboxes
The EvilTokens phishing service, which compromised more than 12,000 inboxes at over 10,000 organizations, has been disrupted by a coalition of law enforcement and private-sector partners led by Microsoft.

WordPress 7.1.2 fixes critical unauthenticated path traversal vulnerability (CVE-2026-87902)
WordPress released version 7.1.2 to fix a critical flaw that lets an unauthenticated attacker make the software load a PHP file of the attacker’s choosing from outside the site’s active theme folders. On sites where the server and the active theme meet certain conditions, the attacker can go on to run code on the server.

Fake Claude Max giveaway tricks users into handing over their Google account credentials
A fake Claude Max giveaway uses a spoofed Google sign-in window to steal users’ login credentials, Malwarebytes researchers have found. “Browser-in-the-browser” is not a new technique. Researchers have documented it since 2022, and in June Palo Alto Networks’ Unit 42 reported a campaign that used draggable fake browser windows to target Microsoft 365 users.

80,000 relay servers help users in China slip past U.S. AI region bans
More than 80,000 relay servers are helping users in China bypass geographic restrictions on leading U.S. AI models, according to Team Cymru. Earlier this month, CISA, the NSA and the FBI warned in a joint advisory that China-based AI firms are running large-scale knowledge distillation campaigns to pull capabilities out of leading U.S. models.

GNOME 50.5 security fixes patch a gvfs CVE and Epiphany code injection
GNOME 50.5, which the GNOME Release Team shipped on September 24, patches a CVE in the gvfs file system layer, a JavaScript injection flaw in the Epiphany web browser and a use-after-free bug in the librsvg image library.

New Android malware RemControl steals banking PINs and blocks removal attempts
A new Android banking trojan called RemControl tricks victims into installing a fake TV app, then takes control of their phones to steal banking PINs, Group-IB has found. Researchers confirmed that the malware targets customers of more than 30 banks in Italy, France, Spain, Poland, Portugal, Canada and some Gulf states.

UK gears up for fight against Russia’s disinformation machine
The UK government will create a new body to track and disrupt disinformation campaigns run by hostile states, Prime Minister Andy Burnham announced at the United Nations General Assembly in New York. In his first address to the Assembly on 22 September, Burnham told world leaders he was tasking UK security chiefs to begin work on the National Centre for Information Defence, which will operate “to detect, attribute and disrupt these kinds of hostile state information attacks.”

Fake payroll desktop apps hand attackers a route to company paychecks
An attacker has been offering “desktop apps” for three large US payroll and HR platforms that have never released one, Allure Security have found. Anyone who runs the installer gets a copy of ScreenConnect, a legitimate remote access tool, configured to let the attacker control the computer without the user knowing.

MacSync info-stealing malware hides malicious commands in an iCloud calendar
A new MacSync variant targets Mac users with an infostealer and persistent backdoor designed to steal credentials, crypto wallet data, and files, according to Kaspersky. Researchers found the malware spreading through a crypto wallet app called Toria, which had its own website and was promoted on X and Telegram.

Threat detection dashboards are masking security coverage gaps
A detection rule can show up as deployed on a coverage dashboard and still never fire when an attacker uses the technique it was built to catch. Conifers assessed 14,652 detections in its customer base, including rules written by customers and detections managed by vendors in SIEM, endpoint, cloud, identity, email and network tools. The research found that 47% of detections in the average organization need attention.

AI compliance issues hit 2 in 5 large companies, and legacy workflows are a big factor
Forty percent of large companies had an AI-related compliance or governance issue in the past 12 months, according to 1,000 senior IT, operations, and transformation leaders surveyed by Sapio Research. Those leaders said process-related problems contributed to 84 percent of the incidents.

Gopass: Open-source command-line password manager for teams
Gopass is a free, open-source password manager that stores credentials in an encrypted store and runs from the command line. Its maintainers built it as a drop-in replacement for pass, the standard Unix password manager.

Product showcase: Helmit alerts parents when online conversations show signs of trouble
Helmit is a parental control app that combines AI-powered social media monitoring with screen time management, web filtering, location tracking, and safety alerts. It identifies potentially concerning interactions and surface the messages associated with an alert.

Google hit with €403 million GDPR fine over location tracking
Ireland’s Data Protection Commission (DPC) has fined Google €403 million (about $463 million) over its processing of users’ location data and ordered the company to bring that processing into compliance within six months.

DavMail 7.0.0 puts most of its work into Microsoft Graph
Anyone who wants to leave Outlook but still has a mailbox on Exchange needs a translator. DavMail is one: a Java gateway that converts the open protocols most mail, calendar and contact apps speak (IMAP, SMTP, CalDAV, CardDAV and LDAP) into requests Exchange and Office 365 accept. “Ever wanted to get rid of Outlook?” the project page asks.

A cheap fake base station can still track 5G subscribers
Researchers from the i2CAT Foundation, the University of Murcia, and NEC Laboratories Europe built a low-cost tool called 5G-Shark that lures a target phone onto a fake base station and questions it, then used it to audit commercial 5G networks.

The next intellectual property thief may sound like your CEO
Impersonation, phishing and domain-name abuse are the most concerning types of online intellectual property infringement, according to CSC’s The State of Online IP Risk 2026 report. Internet and branded content, online marketplaces and paid search were the channels most frequently targeted.

NetBSD 10.2 security fixes close a remote kernel bug in ipfilter
A NetBSD box at the edge of a network, filtering traffic with ipfilter, has been carrying a kernel flaw that someone outside the machine can set off. The bug is a remotely triggerable null pointer dereference in ipfilter, meaning the kernel tries to read memory through a pointer that leads nowhere. In kernel code, that usually ends with the whole system going down.

Nearly two-thirds of tested websites fail every bot test
Malicious bot activity increased 124% between July 2025 and June 2026, compared with 13.2% growth in human traffic. Traffic from AI agents and large language model crawlers rose 82.3% during the same period, according to DataDome’s State of Bot & Agent Security Report 2026.

Product showcase: Scamwise checks the red flags before you take the bait
Scamwise is a free scam-checking service from Savi that examines suspicious messages, emails, websites, phone numbers, images, and real-world situations for signs of fraud. The service works in any web browser on desktop, mobile, or tablet, with no account required.

Prismor: Open-source runtime control plane for AI agents
Prismor is a free, open-source security layer for AI coding agents. It sits between an agent such as Claude Code, Codex, or Cursor and the actions that agent wants to take, and it checks each tool call against a policy before the call runs. Every call gets one of three verdicts: allow, warn, or block.

Claude Opus 5.5 cuts costs and adds safeguards for autonomous AI
Claude Opus 5.5 is available across Anthropic’s platforms, Amazon Web Services, Google Cloud and Microsoft Azure. Developers can access it through the Claude Platform using the model name claude-opus-5-5. It includes watermarking measures designed to comply with the EU AI Act.

GPT-6 Sol and Luna arrive with 50% lower API prices
OpenAI has expanded GPT-6 with the GPT-6 Sol and GPT-6 Luna models. Both are available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users. Free and Go users can access GPT-6 Luna in the desktop app. The models are not yet available in Chat. OpenAI API users can access them as gpt-6-sol and gpt-6-luna.

Americans’ views on data centers have turned more negative
American attitudes toward data centers have turned noticeably more negative over the course of 2026, according to a new Pew Research Center survey. 54% of U.S. adults now say data centers are mostly bad for the environment, up from 39% in January.

Europe’s technology backbone is becoming a cyber target
Disruptive attacks on public-facing services, financially motivated cybercrime and compromises of shared technology providers are increasing cybersecurity risks across Europe. ENISA’s Threat Landscape 2026 identifies cybercrime, state-linked activity, foreign information manipulation and interference, hacktivism and vulnerability exploitation as key threats.

Ubuntu kernel CVE fixes are moving to a weekly release schedule
Ubuntu kernels will ship every week under a new release schedule from Canonical, which is merging its four-week cycle for regular Stable Release Updates (SRUs) and its two-week cycle for security fixes into a single two-week cycle. The cycles overlap, each starting a week after the one before, which is what produces a weekly release.

Your security program knows about the firewall, but does it know about the elevator?
By early August, attackers had hit water systems in at least seven U.S. states. The FBI and EPA said the intruders remotely accessed internet-facing programmable logic controllers, the small industrial computers that run pumps and valves. Operators lost monitoring or control, and in some cases water operations degraded. Federal investigators are examining possible links to Iran-backed hackers.

Claude.ai is about 3x faster after 3,000+ changes
Anthropic engineers made claude.ai and the Claude desktop app roughly three times faster during a two-week sprint in August, with Claude finding the bottlenecks and writing the fixes. The team merged more than 3,000 changes and says none of them caused a customer-facing incident or rollback.

Apple’s new iOS 27 feature looks for signs you’re being scammed
Apple introduced a scam-prevention feature called Impersonation Risk Detection with iOS 27 and iPadOS 27. The feature allows supported apps to request a risk assessment when a user takes an action that could be connected to an active social engineering scam.

Google plans to give Private AI Compute a memory that follows users across devices
Google plans to add private, server-side memory to Private AI Compute, enabling AI assistants to maintain continuity across devices while providing privacy protections normally associated with on-device processing.

Meta locks itself out of user data on its AI glasses
Meta is expanding Private Processing to its AI glasses, extending their security protections into cloud data centers. The system runs AI models inside confidential virtual machines (CVMs) designed to prevent Meta from accessing users’ data. Private Processing combines protected hardware, encryption and software verification to secure data during cloud processing and storage.

Your incident count is missing a few incidents
If you run security for a brand with hundreds or thousands of locations, the tools you’ve bought may have little to do with whether an attack stays at one store. A new VikingCloud survey asked 200 security and IT leaders at U.S. and European chains about the past year. None of the 13 security technologies it measured was tied to less spread between sites, and neither was real-time visibility. One policy decision was.

Half of threat hunters say bad data is their biggest problem
Half of security professionals name data quality or quantity as their biggest barrier to effective threat hunting, according to the SANS 2026 Threat Hunting Survey. Teams with working playbooks describe the logs as their ceiling, and gaps in cloud logging and identity telemetry come up most often.

Cybersecurity jobs available right now: September 22, 2026
We’ve scoured the market to bring you a selection of roles that span various skill levels within the cybersecurity field. Check out this weekly selection of cybersecurity jobs available right now.

New infosec products of the month: September 2026
Here’s a look at the most interesting products from the past week, featuring releases from Akeyless, Akuity, Bitsight, BugBase, Cloud Range, Cohesity, Dataminr, Gurucul, Nozomi Networks, Orchid Security, Ping Identity, Scytale, Securin, Superna, and Tuskira.