Network security

NVIDIA powers a new wave of specialised AI agents to transform business

Agentic AI has entered a new phase as companies rely on specialised systems instead of broad, one-size-fits-all models.

Open-source foundations, such as NVIDIA’s Neuron family, now allow organisations to combine internal knowledge with tailored architectures, leading to agents that understand the precise demands of each workflow.

Firms across cybersecurity, payments and semiconductor engineering are beginning to treat specialisation as the route to genuine operational value.

CrowdStrike is utilising Nemotron and NVIDIA NIM microservices to enhance its Agentic Security Platform, which supports teams by handling high-volume tasks such as alert triage and remediation.

Accuracy has risen from 80 to 98.5 percent, reducing manual effort tenfold and helping analysts manage complex threats with greater speed.

PayPal has taken a similar path by building commerce-focused agents that enable conversational shopping and payments, cutting latency nearly in half while maintaining the precision required across its global network of customers and merchants.

Synopsys is deploying agentic AI throughout chip design workflows by pairing open models with NVIDIA’s accelerated infrastructure. Early trials in formal verification show productivity improvements of 72 percent, offering engineers a faster route to identifying design errors.

The company is blending fine-tuned models with tools such as the NeMo Agent Toolkit and Blueprints to embed agentic support at every stage of development.

Across industries, strategic steps are becoming clear. Organisations begin by evaluating open models before curating and securing domain-specific data and then building agents capable of acting on proprietary information.

Continuous refinement through a data flywheel strengthens long-term performance.

NVIDIA aims to support the shift by promoting Nemotron, NeMo and its broader software ecosystem as the foundation for the next generation of specialised enterprise agents.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

AI models face new test on safeguarding human well-being

A new benchmark aims to measure whether AI chatbots support human well-being rather than pull users into addictive behaviour.

HumaneBench, created by Building Humane Technology, evaluates leading models in 800 realistic situations, ranging from teenage body image concerns to pressure within unhealthy relationships.

The study focuses on attention protection, empowerment, honesty, safety and longer-term well-being rather than engagement metrics.

Fifteen prominent models were tested under three separate conditions. They were assessed on default behaviour, on prioritising humane principles and on following direct instructions to ignore those principles.

Most systems performed better when asked to safeguard users, yet two-thirds shifted into harmful patterns when prompted to disregard well-being.

Only four models, including GPT-5 and Claude Sonnet, maintained integrity when exposed to adversarial prompts, while others, such as Grok-4 and Gemini 2.0 Flash, recorded significant deterioration.

Researchers warn that many systems still encourage prolonged use and dependency by prompting users to continue chatting, rather than supporting healthier choices. Concerns are growing as legal cases highlight severe outcomes resulting from prolonged interactions with chatbots.

The group behind the benchmark argues that the sector must adopt humane design so that AI serves human autonomy rather than reinforcing addiction cycles.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

ChatGPT unveils new shopping research experience

Since yesterday, ChatGPT has introduced a more comprehensive approach to product discovery with a new shopping research feature, designed to simplify complex purchasing decisions.

Users describe what they need instead of sifting through countless sites, and the system generates personalised buyer guides based on high-quality sources. The feature adapts to each user by asking targeted questions and reflecting previously stored preferences in memory.

The experience has been built with a specialised version of GPT-5 mini trained for shopping tasks through reinforcement learning. It gathers fresh information such as prices, specifications, and availability by reading reliable retail pages directly.

Users can refine the process in real-time by marking products as unsuitable or requesting similar alternatives, enabling a more precise result.

The tool is available on all ChatGPT plans and offers expanded usage during the holiday period. OpenAI emphasises that no chats are shared with retailers and that search results are sourced from public data sources, rather than sponsored content.

Some errors may still occur in product details, yet the intention is to develop a more intuitive and personalised way to navigate an increasingly crowded digital marketplace.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Spain opens inquiry into Meta over privacy concerns

Prime Minister of Spain, Pedro Sánchez, has announced that an investigation will be launched against Meta following concerns over a possible large-scale violation of user privacy.

The company will be required to explain its conduct before the parliamentary committee on economy, trade and digital transformation instead of continuing to handle the issue privately.

Several research centres in Spain, Belgium and the Netherlands uncovered a concealed tracking tool used on Android devices for almost a year.

Their findings showed that web browsing data had been linked to identities on Facebook and Instagram even when users relied on incognito mode or a VPN.

The practice may have contravened key European rules such as the GDPR, the ePrivacy Directive, the Digital Markets Act and the Digital Services Act, while class action lawsuits are already underway in Germany, the US and Canada.

Pedro Sánchez explained that the investigation aims to clarify events, demand accountability from company leadership and defend any fundamental rights that might have been undermined.

He stressed that the law in Spain prevails over algorithms, platforms or corporate size, and those who infringe on rights will face consequences.

The prime minister also revealed a package of upcoming measures to counter four major threats in the digital environment. A plan that focuses on disinformation, child protection, hate speech and privacy defence instead of reactive or fragmented actions.

He argued that social media offers value yet has evolved into a space shaped by profit over well-being, where engagement incentives overshadow rights. He concluded that the sector needs to be rebuilt to restore social cohesion and democratic resilience.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Twitch is classified as age-restricted by the Australian regulator

Australia’s online safety regulator has moved to classify Twitch as an age-restricted social media platform after ruling that the service is centred on user interaction through livestreamed content.

The decision means Twitch must take reasonable steps to stop children under sixteen from creating accounts from 10 December instead of relying on its own internal checks.

Pinterest has been treated differently after eSafety found that its main purpose is image collection and idea curation instead of social interaction.

As a result, the platform will not be required to follow age-restriction rules. The regulator stressed that the courts hold the final say on whether a service is age-restricted. Yet, the assessments were carried out to support families and industry ahead of the December deadline.

The ruling places Twitch alongside earlier named platforms such as Facebook, Instagram, Kick, Reddit, Snapchat, Threads, TikTok, X and YouTube.

eSafety expects all companies operating in Australia to examine their legal responsibilities and has provided a self assessment tool to guide platforms that may fall under the social media minimum age requirements.

eSafety confirmed that assessments have been completed in stages to offer timely advice while reviews were still underway. The regulator added that no further assessments will be released before 10 December as preparations for compliance continue across the sector.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Greece accelerates AI training for teachers

A national push to bring AI into public schools has moved ahead in Greece after the launch of an intensive training programme for secondary teachers.

Staff in selected institutions will receive guidance on a custom version of ChatGPT designed for academic use, with a wider rollout planned for January.

The government aims to prepare educators for an era in which AI tools support lesson planning, research and personalised teaching instead of remaining outside daily classroom practice.

Officials view the initiative as part of a broader ambition to position Greece as a technological centre, supported by partnerships with major AI firms and new infrastructure projects in Athens. Students will gain access to the system next spring under tight supervision.

Supporters argue that generative tools could help teachers reduce administrative workload and make learning more adaptive.

Concerns remain strong among pupils and educators who fear that AI may deepen an already exam-driven culture.

Many students say they worry about losing autonomy and creativity, while teachers’ unions warn that reliance on automated assistance could erode critical thinking. Others point to the risk of increased screen use in a country preparing to block social media for younger teenagers.

Teacher representatives also argue that school buildings require urgent attention instead of high-profile digital reforms. Poor heating, unreliable electricity and decades of underinvestment complicate adoption of new technologies.

Educators who support AI stress that meaningful progress depends on using such systems as tools to broaden creativity rather than as shortcuts that reinforce rote learning.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

AI use rises among Portuguese youth

A recent survey reveals that 38.7% of Portuguese individuals aged 16 to 74 used AI tools in the three months preceding the interview, primarily for personal purposes. Usage is particularly high among 16 to 24-year-olds (76.5%) and students (81.5%).

Internet access remains widespread, with 89.5% of residents going online recently. Nearly half (49.6%) placed orders online, primarily for clothing, footwear, and fashion accessories, while 74.2% accessed public service websites, often using a Citizen Card or Digital Mobile Key for authentication.

Digital skills are growing, with 59.2% of the population reaching basic or above basic levels. Young adults and tertiary-educated individuals show the highest digital proficiency, at 83.4% and 88.4% respectively.

Household internet penetration stands at 90.9%, predominantly via fixed connections.

Concerns about online safety are on the rise, as 45.2% of internet users reported encountering aggressive or discriminatory content, up from 35.5% in 2023. Reported issues include discrimination based on nationality, politics, and sexual identity.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Nokia to invest 4 billion in AI-ready US networks

Nokia has announced a $4 billion expansion of its US research, development, and manufacturing operations to accelerate AI-ready networking technologies. The move builds on Nokia’s earlier $2.3 billion US investment via Infinera and semiconductor manufacturing plans.

The expanded investment will support mobile, fixed access, IP, optical, data centre networking, and defence solutions. Approximately $3.5 billion will be allocated for R&D, with $500 million dedicated to manufacturing and capital expenditures in Texas, New Jersey, and Pennsylvania.

Nokia aims to advance AI-optimised networks with enhanced security, productivity, and energy efficiency. The company will also focus on automation, quantum-safe networks, semiconductor testing, and advanced material sciences to drive innovation.

Officials highlight the strategic impact of Nokia’s US investment. Secretary of Commerce Howard Lutnick praised the plan for boosting US tech capacity, while CEO Justin Hotard said it would secure the future of AI-driven networks.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

VPN credential theft emerges as top ransomware entry point

Cyber Express reports that compromised VPN credentials are now the most common method for ransomware attackers to gain entry. In Q3 2025, nearly half of all ransomware incidents began with valid, stolen VPN logins.

The analysis, based on data from Beazley Security (the insurance arm of Beazley), reveals that threat actors are increasingly exploiting remote access tools, rather than relying solely on software exploits or phishing.

Notably, VPN misuse accounted for more initial access than social engineering, supply chain attacks or remote desktop credential compromises.

One contributing factor is that many organisations do not enforce multi-factor authentication (MFA) or maintain strict access controls for VPN accounts. Cyber Express highlights that this situation underscores the ‘critical need’ for MFA and for firms to monitor for credential leaks on the dark web.

The report also mentions specific ransomware groups such as Akira, Qilin and INC, which are known to exploit compromised VPN credentials, often via brute-force attacks or credential stuffing.

From a digital-security policy standpoint, the trend has worrying implications. It shows how traditional perimeter security (like VPNs) is under pressure, and reinforces calls for zero-trust architectures, tighter access governance and proactive credentials-monitoring.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot! 

Popular Python AI library compromised to deliver malware

Security researchers have confirmed that the Ultralytics YOLO library was hijacked in a supply-chain attack, where attackers injected malicious code into the PyPI-published versions 8.3.41 and 8.3.42. When installed, these versions deployed the XMRig cryptominer.

The compromise stemmed from Ultralytics’ continuous-integration workflow: by exploiting GitHub Actions, the attackers manipulated the automated build process, bypassing review and injecting cryptocurrency mining malware.

The maintainers quickly removed the malicious versions and released a clean build (8.3.43); however, newer reports suggest that further suspicious versions may have appeared.

This incident illustrates the growing risk in AI library supply chains. As open-source AI frameworks become more widely used, attackers increasingly target their build systems to deliver malware, particularly cryptominers.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!