Sign in Subscribe

InfoSec News 18MAR2026

General

  • Microsoft are having more issues with classic Outlook - this time it's the Teams plugin.
Microsoft is working to address a known issue that renders the classic Outlook email client unusable for users who have enabled the Microsoft Teams Meeting Add-in.
"Some users may be unable to use Microsoft Outlook Classic while the Microsoft Teams Meeting Add-in is enabled," Microsoft said in an admin center update under EX1254044.
While it didn't share more details about the root cause, Microsoft says the issue is due to a previous Outlook build.
...
Microsoft is also investigating several issues that cause email connection problems in the classic Outlook desktop client and trigger 0x800CCC0F and 0x80070057 errors during synchronization with Gmail and Yahoo accounts.
In January, it addressed another classic Outlook issue caused by the December 2025 updates that prevented Microsoft 365 customers from opening encrypted emails.
On Monday, Redmond also mitigated a widespread Exchange Online outage (EX1253275) that prevented customers from accessing their mailboxes and calendars via Outlook on the web, Outlook desktop, Exchange ActiveSync, and other Exchange Online connection protocols.
Microsoft has released an emergency update to fix a Bluetooth device visibility issue on hotpatch-enabled Windows 11 Enterprise devices.
The Bluetooth devices will not appear in Windows Settings or Quick Settings, even though they're functioning correctly and connected to affected systems.
This bug could also prevent some users from adding new Bluetooth devices because available devices will not appear in the connection list.
On March 10, 2026, Zhou Hongyi announced 360 Security Lobster (360安全龙虾), a commercial wrapper around the open-source AI agent OpenClaw. The product was positioned as a solution to OpenClaw's three primary problems: high installation barriers (usage threshold too high), unpredictable results (results too random), and security vulnerabilities (security risks too prominent). Zhou described OpenClaw as "a remarkable innovation" but likened it to "an intern" that requires patient training. 360 Security Lobster was framed as the enterprise-grade fix, reducing setup time from approximately six hours to ten minutes.
...
On March 16, 2026, security researchers discovered that the installer package contained the wildcard SSL private key for `*.myclaw.360.cn`
...
The leaked certificate was issued by WoTrus CA Limited. WoTrus is the rebranded version of WoSign CA Limited, a Chinese certificate authority. Qihoo 360 owns WoSign and, by extension, WoTrus.
...
In 2015, WoSign secretly acquired StartCom, an Israeli certificate authority founded in 1999, without disclosing the deal to browser vendors who operate certificate root programs. WoSign and StartCom repeatedly denied the acquisition until Mozilla gathered sufficient evidence proving otherwise. The two CAs were found to be sharing infrastructure, staff, policies, and issuance systems while presenting themselves as separate entities.

Getting Techy

  • LeakNet have a new strategy to help evade detection. Much like previous cases of calling PowerShell directly with attacker code (not pointing to a file), LeakNet are using the Deno JavaScript runtime. This will likely be less instrumented in environments, however they do need to install it on the systems first. The choice of Deno is somewhat 'interesting' - Deno is designed to be a "Secure by default" JavaScript/TypeScript runtime. As its website states "A program run with Deno has no file, network, or environment access unless explicitly enabled."
Deno is a legitimate runtime developers use to execute JavaScript and TypeScript. LeakNet abuses that legitimacy with a “bring your own runtime” (BYOR) approach. Rather than deploying a custom malware loader that’s more likely to get flagged, the attackers install the legitimate Deno executable and use it to run malicious code. In observed activity, that process was initiated through Visual Basic Script (VBS) and PowerShell scripts, cleverly named Romeo*.ps1 and Juliet*.vbs.
The key advantage is delivery. Instead of writing a malicious JavaScript file to disk, LeakNet passes the code to Deno as a base64-encoded data: URL. Deno then decodes and executes it directly in memory. In practice, that means the script never appears as a normal file on the endpoint, reducing forensic artifacts and making the activity look more like a one-off developer task than a traditional C2 loader.
Deno also provides a --allow-all flag that grants all permissions to the script. This disables the security sandbox entirely, and should be used with caution. The --allow-all has the same security properties as running a script in Node.js (ie none).
Definition: -A, --allow-all
SCION, which stands for Scalability, Control, and Isolation On Next-Generation Networks, is an internet routing architecture developed at ETH Zürich. Unlike the patches applied to BGP, SCION does not attempt to retrofit security onto a 40-year-old foundation. It replaces the foundation entirely. That redesign is the life's work of Adrian Perrig, professor of computer science at ETH Zürich and the principal architect of SCION.
Ctrl-Alt-Intel linked this infrastructure back to CERT-UA reporting from September 2024, meaning FancyBear operated from the same C2 server for over 500 days, long after it was publicly attributed to GRU-linked operations. Despite being classified as a “sophisticated” threat, FancyBear demonstrated poor operational security on multiple counts, giving us rare, near-complete visibility into Russian espionage operations.

Geo-Politics

  • [EU] The European Council / Council of the European Union has sanctioned two chinese companies and an Iranian individual over cyber-attacks on the EU.
The Council has listed Integrity Technology Group, a China-based company, that has routinely provided products used to compromise and access devices in EU members states, across Europe and worldwide. Between 2022 and 2023, through their technical and material support, more than 65,000 devices were hacked across six member states.
Similarly, Anxun Information Technology, a China-based company, has provided hacking services aimed at the critical infrastructure and critical functions of member states and third countries. The two Chinese individuals also listed today by the Council, are co-founders of the company and were responsible for and involved in cyber-attacks affecting EU member states.
Lastly, the Iranian company Emennet Pasargad has unlawfully gained access to a French subscriber database and advertised its contents for sale on the dark web. They also compromised advertising billboards to spread disinformation during the 2024 Paris Olympic Games. Additionally, the company compromised a Swedish SMS service, impacting a large number of EU citizens.
Per reports, at least two Iranians accused of running cyber operations against Western entities were killed in the strikes. One was Mohammad Mehdi Farhadi Ramin, who the Justice Department charged in 2020 with hacking into aerospace and defense companies in America and had been sought by U.S. authorities since then.
The other was Seyed Yahya Hosseiny Panjaki, a deputy minister of intelligence for Israeli affairs at MOIS who is on the FBI Most Wanted list. Cybersecurity sources tell Forbes Panjaki was in charge of the MOIS unit that controlled hacking groups like Handala, long known as a pro-Iranian crew that has successfully targeted Israeli politicians and Western businesses. According to the FBI, Panaki was linked to terror plots and cyberattacks organized by MOIS and and Iran's Islamic Revolutionary Guard Corps.”

Privacy

  • [US] Retailer Sears' chat bot was storing conversations in an unsecured database.
The exposed Sears databases uncovered by Fowler, which have since been secured, contained 3.7 million chat logs, plus 1.4 million audio files and plain text transcripts from 2024 to this year. Fowler found that one CSV file about the incident contained 54,359 complete chat logs. Conversations Fowler saw included the chatbot introducing itself as “Samantha, an AI virtual voice agent for Sears Home Services,” with the logs also including the name of the company’s AI technology “kAIros.” The cache of data contained chats in both English and Spanish and included personal information about Sears customers, such as names, phone numbers, home addresses, appliances owned, and information on delivery appointments and repairs.

AI

  • Mistral have released a new open-weights model - Mistral Small 4 - designed to combine all of the capabilities of their previous models, in a 119b-parameter, Mixture of Experts (MoE) model with 6b active. 256k context window.
Today, we are announcing Mistral Small 4. This model is the next major release in the Mistral Small family. Mistral Small 4 is the first Mistral model to unify the capabilities of our flagship models, Magistral for reasoning, Pixtral for multimodal, and Devstral for agentic coding, into a single, versatile model. With Small 4, users no longer need to choose between a fast instruct model, a powerful reasoning engine, or a multimodal assistant: one model now delivers all three, with configurable reasoning effort and best-in-class efficiency.
Mistral Small 4 is released under the Apache 2.0 license, continuing our commitment to open, accessible, and customizable AI.
...
Mistral Small 4 is a hybrid model optimized for general chat, coding, agentic tasks, and complex reasoning. Its architecture supports both text and image inputs, making it versatile for a wide range of applications.
...
Mixture of Experts (MoE): 128 experts, with 4 active per token, enabling efficient scaling and specialization.
119B total parameters, with 6B active parameters per token (8B including embedding and output layers).
256k context window, supporting long-form interactions and document analysis.
Configurable reasoning effort: Toggle between fast, low-latency responses and deep, reasoning-intensive outputs.
Native multimodality: Accepts both text and image inputs, unlocking use cases from document parsing to visual analysis.
...
40% reduction in end-to-end completion time (latency-optimized setup).
3x more requests per second (throughput-optimized setup) compared to Mistral Small 3.
Today we’re releasing GPT‑5.4 mini and nano, our most capable small models yet. They bring many of the strengths of GPT‑5.4 to faster, more efficient models designed for high-volume workloads.
GPT‑5.4 mini significantly improves over GPT‑5 mini across coding, reasoning, multimodal understanding, and tool use, while running more than 2x faster. It also approaches the performance of the larger GPT‑5.4 model on several evaluations, including SWE-Bench Pro and OSWorld-Verified.
GPT‑5.4 nano is the smallest, cheapest version of GPT‑5.4 for tasks where speed and cost matter most. It is also a significant upgrade over GPT‑5 nano. We recommend it for classification, data extraction, ranking, and coding subagents that handle simpler supporting tasks.
...
GPT‑5.4 mini is also a strong fit for systems that combine models of different sizes. In Codex, for example, a larger model like GPT‑5.4 can handle planning, coordination, and final judgment, while delegating to GPT‑5.4 mini subagents that handle narrower subtasks in parallel—like searching a codebase, reviewing a large file, or processing supporting documents.
...
In the API, GPT‑5.4 mini supports text and image inputs, tool use, function calling, web search, file search, computer use, and skills. It has a 400k context window and costs $0.75 per 1M input tokens and $4.50 per 1M output tokens.
...
GPT‑5.4 nano is only available in the API and costs $0.20 per 1M input tokens and $1.25 per 1M output tokens.
Microsoft has stopped automatically installing the Microsoft 365 Copilot app on Windows devices that have the Microsoft 365 desktop client apps.
The Microsoft 365 Copilot app integrates the AI-powered Copilot assistant with Microsoft 365 suite apps, including Word, Excel, and PowerPoint, as well as other features like AI agents and Notebooks.
...
Over the weekend, Windows Central also reported that Microsoft might be quietly canceling plans to ship several other Copilot-powered features introduced by Microsoft EVP Yusef Mehdi almost two years ago, which would have embedded Copilot into the Windows 11 Settings app, system notifications, and the File Explorer.
He raised the possibility of a Friday afternoon AI ban when advising on the fifth risk he has identified: Copilot producing output that is toxic because while it may be factually correct it is culturally unacceptable either in the workplace or among customers. Xu recommended mitigating Copilot’s tendency to produce toxic content by enabling the filters Microsoft supplies, and by training users to always validate the tool’s output.
The analyst reminded the audience that all Copilot output isn’t fit for sharing without review, making validation necessary for all users at all times. He suggested Friday afternoons are a time when workers might just want to get the job done and won’t bother to check for errors that Microsoft’s chatbot produces, perhaps making that slice of the working week a fine time to ban use of Copilot.
...
“Copilot makes over-shared documents more accessible,” he warned. “This is not a net new risk, but a known risk amplified by AI.” Xu explained why with the example of a worker who uses Copilot to search for information about organizational changes receiving a response that includes a confidential document about an imminent re-org.
World, known as WorldCoin until late 2024 when AI became trendier than cryptocurrency, announced on Tuesday that it was opening a limited beta of its new AgentKit. The new tech, says World, will serve as a way to tie AI agents directly to a human to prevent bad actors from abusing agentic AI and "infuse trust into the system."
...
AgentKit serves as an extension to Coinbase's x402 protocol, which allows cryptocurrency users to exchange digital cash directly over HTTP.
x402 has also been extended to serve as a way to limit AI agent access to online resources by charging them micropayments. While that's enough to filter out some bad actors, World argues that actual identity verification is needed.
"Through World ID, a person can cryptographically and anonymously prove that they are a unique human without revealing any personal information to anyone," World said in a press release. "That same proof can now extend to their agents."
A judge ordered the reinstatement of a video game developer after he was fired as part of a scheme cooked up by a CEO using ChatGPT. Facing the possibility of paying out a massive bonus to the developer of Subnautica 2, the CEO of publisher Krafton used ChatGPT to create a plan to take over the development studio and force out its founder, according to court records.
...
Kim partnered with Krafton Head of Corporate Development Maria Park and the company’s legal team to work out options. He toyed with finding a reason to fire the founders. According to court records, Park pinged Kim on Slack and told him that attempting to avoid paying the bonus would be legally risky. “Hi CEO . . . it seems to be highly likely that the earn-out will still be paid if the sales goal is achieved regardless of the dismissal with cause,” the Slack message said according to court records. “Therefore, there isn’t much that we can practically gain other than punishment with a simple dismissal alone, whereas I am worried that we may be exposed to lawsuit and reputation risk.”
But the CEO would not accept defeat. “And so Kim turned to ChatGPT for help,” court records said.

Subscribe to Deuxieme RE Banque News

Sign up now to get access to the library of members-only issues.
Jamie Larson
Subscribe