Security

BADCANDY Hack Exploits Cisco IOS XE, Warns ASD

Insights Desk

A bulletin about continuing cyberattacks targeting unpatched Cisco IOS XE equipment in the nation using an unauthorized implant known as BADCANDY has been released by the Australian Signals Directorate (ASD).

According to the intelligence agency, the activity entails the exploitation of CVE-2023-20198 (CVSS score: 10.0), a critical vulnerability that enables a remote, unauthenticated attacker to create an account with elevated privileges and use it to take over vulnerable systems.

Since 2023, the security flaw has been actively exploited in the open; in recent months, threat actors connected to China, such as Salt Typhoon, have used it as a weapon to compromise telecom companies.

Variations of BADCANDY have been identified since October 2023, according to ASD, with new attacks being reported in 2024 and 2025. The malware is thought to have infected up to 400 devices in Australia since July 2025, with 150 of those devices being affected in October alone.

It cannot endure system reboots because it lacks a persistence mechanism. However, the threat actor may re-introduce the malware and recover access if the device is left unpatched and connected to the internet.

According to ASD’s assessment, threat actors can recognize when an implant is removed and re-infect the devices. This is because devices for which the agency has previously notified impacted entities have been re-exploited.

Nevertheless, a reboot won’t reverse the attackers’ previous actions. Therefore, in order to stop further exploitation attempts, system operators must apply the patches, restrict the web user interface’s public exposure, and adhere to Cisco’s hardening standards.

Security

Meta Rolls Out Easier Encryption for WA Cloud Backups

Insights Desk

With passkey support, WhatsApp is introducing a new method for accessing your encrypted backups. This implies that you can access WhatsApp’s backup in the event that you misplace your handset using techniques like fingerprint, facial recognition, or the screen lock code of your old device.

WhatsApp’s chat backups lacked an encryption layer for many years. But in 2021, Meta introduced a feature that allows customers to secure their backups with end-to-end encryption using 64-character encryption keys or passwords.

The issue with both is that, to restore the backup, you must either remember your backup password or have the encryption key on hand. Users don’t have to search for the key or the password while using passkeys.

They may need to monitor when this function is made available, as WhatsApp, which surpassed 3 billion active users in May, stated that it will be made available to users in the upcoming weeks and months.

To enable encrypted backups and see if users have the option to utilize passkeys, it can be navigated to Settings > Chats > Chat backup > End-to-end encrypted backup.

Artificial Intelligence

IBM Partners with Groq to Boost AI Speed and Scalability

Insights Desk

Groq and IBM established a technology and go-to-market partnership. Through this partnership, clients may quickly utilize Groq’s GroqCloud inference technology with Watsonx Orchestrate, offering quick and inexpensive AI inference and hastening the deployment of agent-based AI.

The two businesses intend to combine Groq’s LPU architecture with Red Hat’s open source vLLM technology as part of their alliance.

For IBM clients, GroqCloud will also support the IBM Granite model. Even while many businesses are transitioning AI agents from pilot programs to production settings, there are still issues with speed, cost, and dependability, especially in mission-critical sectors like manufacturing, healthcare, finance, government, and retail.

This partnership gives companies the infrastructure they need to succeed by combining IBM’s agent-based AI orchestration with Groq’s inference speed, affordability, and access to the newest open source models. GroqCloud provides inference that is more than five times faster and more economical than conventional GPU systems thanks to its special LPU.

This ensures dependable performance and constantly low latency even as workloads grow globally. For agent-based AI in regulated industries, this is a particularly potent capacity. For instance, thousands of intricate patient inquiries are received concurrently by IBM’s healthcare clients.

IBM’s AI agents can evaluate data in real time and give prompt, accurate responses thanks to Groq’s inference technology, which enhances customer satisfaction and facilitates quicker, more intelligent decision-making. Non-regulated sectors are also using Groq’s inference technology.

IBM’s retail and consumer goods clients are already using it as an HR agent to increase employee productivity and better automate HR procedures.

“Large enterprises have many options when experimenting with AI inferencing, but when moving to production, they need to confidently deploy complex workflows and ensure a high-quality experience,”

said Rob Thomas, Senior Vice President and Chief Commercial Officer, IBM Software.

“Our partnership with Groq underscores IBM’s commitment to providing clients with cutting-edge technology for AI adoption and business value creation.”

Artificial Intelligence

NotebookLM Gets Major Boost in Context and Customization

Insights Desk

Google has rolled out a sweeping update to its AI-powered research and note-taking tool, NotebookLM, introducing powerful new capabilities aimed at boosting performance, personalization, and productivity.

The latest release brings a suite of enhancements, including a significantly larger context window, extended conversation memory, and smarter response generation—marking one of the most substantial upgrades to the platform to date.

According to Google, the revamped NotebookLM now features an 8x larger context window and 6x longer conversation memory, allowing users to conduct deeper, more coherent discussions with the system.

The company reports a 50% improvement in response quality, alongside a range of back-end optimizations designed to make interactions faster and more reliable.

One of the most notable updates is the integration of Gemini’s full 1 million-token context window into NotebookLM’s chat feature.

This expansion enables the AI to analyze and synthesize insights from vast document collections with greater precision and contextual awareness. Additionally, the tool can now retain multiturn conversation history, improving the continuity and relevance of extended exchanges.

Google has also enhanced NotebookLM’s ability to surface and synthesize information. The system now automatically explores user sources from multiple perspectives, generating responses that go beyond the original query.

This approach is especially beneficial for users managing large notebooks, where nuanced understanding and context management are essential.

Another major feature is the introduction of goal-setting in Chat, allowing users to define specific objectives that guide the AI’s tone and focus. Each notebook can now adapt dynamically to the user’s unique workflow, making the tool more personalized and purpose-driven.

To support long-term projects, Google has added automatic chat saving, enabling users to pause and resume sessions without losing conversation history. Privacy controls remain intact, with the option to delete chat records anytime, and in shared notebooks, personal chat data stays private.

With these updates, Google aims to make NotebookLM not only more powerful but also more intuitive—helping creators, researchers, and professionals unlock new levels of creativity and efficiency.

Artificial Intelligence

New Accenture Tool Drives Software-Defined Manufacturing

Insights Desk

Accenture introduced “Physical AI Orchestrator,” a technology to assist manufacturers in reimagining their current and future factories and warehouses so they are software-defined.

The cloud-based solution integrates NVIDIA Metropolis, AI agents from Accenture’s AI RefineryTM platform, and NVIDIA Omniverse, including the “Mega” NVIDIA Omniverse Blueprint.

Virtual copies in a software-defined facility mimic the actual automated plant or warehouse and its apparatus. These real-time digital twins identify problems and simulate the effects of possible process modifications using precise physics.

The physical plant can then adjust to shifting demand, quality, or timing thanks to AI agents’ conversion of the observations into precise commands. Manufacturers may create live digital twins of planned and actual physical assets, such as conveyors, industrial and mobile robots, shop floor and warehouse layouts, and link them to their physical counterparts using Accenture’s Physical AI Orchestrator.

For instance, a solution for worker safety in factories and warehouses has been developed by Belden, a provider of network and data solutions. It created a virtual safety fence solution using Physical AI Orchestrator to surround robots with safety zones without interfering with ongoing activities.

The robots are automatically stopped or redirected if a human approaches the area. With centimeter-level fidelity, the system employs edge AI to identify and simulate worker, vehicle, and robot motions as well as equipment pathways inside buildings.

The virtual safety fence, which has been trained on a variety of hypothetical situations, including unexpected forklift reversals, is anticipated to be initially implemented by an automaker to improve pedestrian safety in warehouse settings.

“Physical AI Orchestrator acts as a brain for a physical space,”

said Prasad Satyavolu, Americas Lead of Accenture’s Digital Engineering and Manufacturing Service, Industry X.

“Powered by NVIDIA Omniverse technologies and Accenture AI Refinery, it is designed to enable software-defined factories and to make agentic AI and physical AI part of the fabric of manufacturing. We are already seeing it provide quick and lasting benefits to our clients across the globe. This is particularly relevant to companies in the US, where manufacturing reinvention is a prerequisite for reindustrialization.”

Artificial Intelligence

Two-Thirds of EMEA Firms See AI Productivity Gains

Insights Desk

Businesses in Europe, the Middle East, and Africa (EMEA) are already reporting notable productivity improvements from utilizing AI, according to a recent survey from IBM.

Many of these businesses anticipate returns on their investments (ROI) within the next year. However, the results indicate that when it comes to using AI to increase productivity, public sector organizations and small to medium-sized enterprises (SME) are lagging larger, private sector companies.

According to a recent IBM analysis titled “The Race for ROI,” which was created in collaboration with Census wide and polled 3,500 senior executives in ten different countries, 66% of participants claimed that their companies have significantly increased operational productivity using AI.

A further 42% of respondents on average anticipated achieving ROI within 12 months across cost reduction (41%), time savings (45%), increased revenue (37%), employee satisfaction (42%), and increased Net Promoter Score (43%).

Additionally, roughly one in five respondents stated their organization has already achieved ROI goals from AI-driven productivity initiatives.

With 92% of leaders anticipating that agentic AI would yield quantifiable ROI within two years, further productivity gains are anticipated from the implementation of AI Agents.

The survey found that software development and IT (32%), customer service (32%), and procurement (27%) are the business sectors with the largest AI-driven productivity increases.

The top three advantages of increased productivity, according to executives, were improved decision-making (50%), increased operational efficiency (55%), and increased worker skills including automating repetitive jobs (48%).

But not every kind of organization benefits equally. Only 55% of SMEs reported productivity increases using AI, compared to 72% of large firms surveyed.

According to the study, just 55% of public sector businesses have reported notable productivity gains thus far, suggesting that they are still in the early stages of achieving AI’s full potential.