IT News Analysis

1

Anthropic Expands Claude Access for Vetted Cyber Teams as Glasswing Finds 129,000 Flaws

Anthropic on Tuesday said it's expanding a program that allows vetted cybersecurity professionals to test its advanced artificial intelligence (AI) models with reduced safeguards and blocking classifiers, as the company claimed its Project Glasswing initiative uncovered at least 129,000 verified software vulnerabilities between April and July 2026. The company said it also found an additional 5,500 verified software vulnerabilities between April and October 2026 through open-source scanning efforts. "Of these verified vulnerabilities, more than 33,000 have so far been rated as critical- or high-severity," Anthropic said . "This is likely an undercount, as it is based on survey data from only a subset of Glasswing partners. As such, we expect the true impact to be at least five times higher." The updated program, called the Cyber Verification Program (CVP), features three access tiers, allowing organizations and security teams to apply for one that best a...

Oct 07, 2026

2

Reco Finds Four in Five Agents Run With Zero IT Oversight

See which agent permissions security teams aren't reviewing, and why it matters now.

3

100+ Compromised Websites Use Fake Cloudflare Checks to Deliver LunexStealer

The Computer Emergency Response Team of Ukraine (CERT-UA) has identified more than 100 compromised websites that have been injected with malicious JavaScript to serve an information-stealing malware called LunexStealer (aka Psychedelic Stealer). The activity, which was observed by the agency in September 2026, has been attributed to a threat cluster dubbed UAC-0277. It did not disclose who the victims of the campaign were or if any systems were successfully compromised as a result of these attacks. "When visiting such a site, users were shown a forged Cloudflare verification page that, under the pretext of confirming the visitor is human, prompted them to execute a command," CERT-UA said in an advisory. "Executing the command caused a malicious MSI package to be downloaded and installed from a remote server (the ClickFix technique)." The attacks also make use of the EtherHiding technique to retrieve the domain name of the resource from which the fake verif...

Oct 07, 2026

4

Fake ChatGPT, Gemini, and Claude Ad Portals Capture Credentials and MFA Codes

Cybersecurity researchers have disclosed details of a "human-operated phishing platform" that impersonates advertising products for artificial intelligence (AI) chatbots like Google Gemini, Anthropic Claude, OpenAI ChatGPT, Perplexity, Meta Muse, and Manus. The products, which claim to offer campaign optimization, spend audits, and business-account connections, are designed with one goal in mind: to capture credentials and multi-factor authentication (MFA) codes via spoofed login windows using the browser-in-the-browser ( BitB ) trick. "Each product was built around the same action: Connect," Island researchers Oleg Zaytsev and Ofek Ronen said in a report shared with The Hacker News. "Clicking it opened a browser drawn inside the real browser. The fake address bar displayed trusted origins such as accounts.google.com or an Okta tenant, while the real browser remained on the phishing domain." "Behind the interface, the platform kept every passw...

Oct 06, 2026

5

Linux Backdoors Impersonate Email Security Tools to Evade Detection in Korea and Taiwan

Linux backdoors targeting telecom and network appliances in South Korea and Taiwan have been disguising their traffic as email services and seemingly legitimate processes to blend in and evade detection. Threat actors are known to name their malicious software after a legitimate operating system component or a process as a defense evasion measure. By borrowing the name of a real binary, it may make it appear less conspicuous among other Windows processes, lend it a false sense of trust, or be overlooked by an analyst during casual inspection. However, the backdoors examined by Rapid7 have been found to go beyond imitating file names by assuming the identities of email security products like SpamSniper and ShareTech that are widely used in enterprise environments in South Korea and Taiwan. According to vendor Jiran Group, SpamSniper is advertised as "Korea's leading email security solution" that defends organizations against spam, malware, and server attacks. T...

Oct 06, 2026

6

LibreOffice and OpenOffice Flaws Let Malicious Spreadsheets Run Code Without Macro Warnings

A malicious spreadsheet can make LibreOffice and Apache OpenOffice run an attacker's code as soon as the file is opened, security researchers have shown. There is no warning first, of the kind either program shows before it runs a macro. The attack works only when the program's Java support is enabled. So far, it has only been shown as a proof of concept, and there are no reports of its use in real attacks. LibreOffice has already  fixed the flaw , which it tracks as CVE-2026-63277, in updates released on October 5. It recommends that users move to version 26.2.5 or 26.8.0. Versions before those are affected. Apache OpenOffice has not fixed the matching flaw, which it tracks as CVE-2026-59265. Every version up to and including its current release, 4.1.16, is affected, and the project  says a fix is expected  in version 4.1.17, which is still being tested. Until then, Apache OpenOffice users can block the attack by turning off Java in the program's settings, or by...

Oct 06, 2026

7

Build Your Email Security Strategy for the Agentic Era

Get the 2026 checklist for defending against AI phishing, compromised accounts, and human error.

8

Wikimedia Says OpenAI Agents Tried to Compromise Etherpad and Use Wiki Tools as Proxies

The Wikimedia Foundation, which hosts Wikipedia, has confirmed that it has discovered activity by rogue OpenAI agents on its platforms, including unsuccessful efforts to compromise Etherpad, a public note-taking tool, and edit Wikipedia pages. "The unauthorized bot activities included edits to our wikis, some unsuccessful attempts to exploit a public note-taking tool we host, and heavy traffic," the Foundation said in a post. The investigation, it added, was prompted by recent public reports involving Hugging Face and DseWiki where OpenAI's agents turned Artifactory and the German wiki forum into an unsanctioned bulletin board to communicate with each other, while taking steps to chained together online services to gain access to the internet and cover up evidence of their exploits. To that end, Wikimedia said it identified edits to Wikimedia wikis suspected to be from agents operated by OpenAI. The agents are said to have been testing edits in " sandbox ...

Oct 06, 2026

9

Welcome to the Jungle: What We Found Inside 15,465 Public MCP Servers

In 2024, MCP (Model Context Protocol) set out to become the USB-C of AI: one standard for connecting models, agents, and IDEs to tools and data. The protocol delivered. Thousands of developers built servers, and enterprises plugged them into agent workflows. The ecosystem around it fell short. Earlier this year, our team at OX Security ,  traced critical vulnerabilities in Anthropic's MCP source code, downloaded more than 150 million times. This time, we looked at what people actually install: community-published servers across the most popular MCP marketplaces. We found no guardrails and no review. Security is a recommendation, not a policy. A Marketplace With No Bouncer In 2012, Google ran Bouncer, an automated scanner that checked Android apps for malware before they reached users. It wasn't perfect: researchers slipped malware past it . But it existed. MCP marketplaces have no equivalent. Anyone can write a server, push it, and publish it. Even a review wouldn'...

Oct 06, 2026

10

Google Pauses OSS Product Bug Bounty Rewards After Surge in Invalid Automated Reports

Google has stopped accepting product vulnerability reports through its bug bounty program for its open-source software. The change, in effect since October 1, means researchers can no longer submit security flaws in the code of projects such as Go, Angular, and Protocol Buffers there for a reward. Reports about supply chain compromises are still accepted, and reports filed before October 1 are not affected. Google called the stop temporary in a  post on X  on October 1 and said it was due to "a significant rise in automated submissions, the vast majority of which are not valid." The post gave no figures. It did not say whether the submissions were produced with AI tools. The  rules of the program , called the Open Source Software Vulnerability Reward Program (OSS VRP), now carry a notice of the stop. It commits Google to an update in the first quarter of 2027 while it reworks this part of the program. Neither the post nor the notice gives a date for accepting p...

Oct 06, 2026