cybersecurity

Fake AI assistant steals OpenAI credentials from thousands of Chrome users

A Chrome browser extension posing as an AI assistant has stolen OpenAI credentials from more than 10,000 users. Cybersecurity platform Obsidian identified the malicious software, known as H-Chat Assistant, which secretly harvested API keys and transmitted user data to hacker-controlled servers.

The extension, initially called ChatGPT Extension, appeared to function normally after users provided their OpenAI API keys. Analysts discovered that the theft occurred when users deleted chats or logged out, triggering the transmission of credentials via hardcoded Telegram bot credentials.

At least 459 unique API keys were exfiltrated to a Telegram channel months before they were discovered in January 2025.

Researchers believe the malicious activity began in July 2024 and continued undetected for months. Following disclosure to OpenAI on 13 January, the company revoked compromised API keys, though the extension reportedly remained available in the Chrome Web Store.

Security analysts identified 16 related extensions sharing the identical developer fingerprints, suggesting a coordinated campaign by a single threat actor.

LayerX Security consultant Natalie Zargarov warned that whilst current download numbers remain relatively low, AI-focused browser extensions could rapidly surge in popularity.

The malicious extensions exploit vulnerabilities in web-based authentication processes, creating, as researchers describe, a ‘materially expanded browser attack surface’ through deep integration with authenticated web applications.

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

UK firms prioritise cyber resilience and AI growth

Cybersecurity is set to receive the largest budget increases over the next 12 months, as organisations respond to rising geopolitical tensions and a surge in high-profile cyber-attacks, according to the KPMG Global Tech Report 2026.

More than half of UK firms plan to lift cybersecurity spending by over 10 percent, outpacing global averages and reflecting heightened concern over digital resilience.

AI and data analytics are also attracting substantial investment, with most organisations increasing budgets as they anticipate stronger returns by the end of 2026. Executives expect AI to shift from an efficiency tool to a core revenue driver, signalling a move toward large-scale deployment.

Despite strong investment momentum, scaling remains a major challenge. Fewer than one in 10 organisations report fully deployed AI or cybersecurity systems today, although around half expect to reach that stage within a year.

Structural barriers, fragmented ownership, and unclear accountability continue to slow execution, highlighting the complexity of translating strategy into operational impact.

Agentic AI is emerging as a central focus, with most organisations already embedding autonomous systems into workflows. Demand for specialist AI roles is rising, alongside closer collaboration to ensure secure deployment, governance, and continuous monitoring.

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

Ten cybersecurity predictions for 2026 from experts: How AI will reshape cyber risks

Evidence from threat intelligence reporting and incident analysis in 2025 suggests that AI will move from experimental use to routine deployment in malicious cyber operations in 2026. Rather than introducing entirely new threats, AI is expected to accelerate existing attack techniques, reduce operational costs for attackers, and increase the scale and persistence of campaigns.

Security researchers and industry analysts point to ten areas where AI is most likely to reshape the cyber threat landscape over the coming year:

  1. AI-enabled malware is expected to adapt during execution. Threat intelligence reporting indicates that malware using AI models is already capable of modifying behaviour in real time. In 2026, such capabilities are expected to become more common, allowing malicious code to adjust tactics in response to defensive measures.
  2. AI agents are likely to automate key stages of cyberattacks. Researchers expect wider use of agentic AI systems that can independently conduct reconnaissance, exploit vulnerabilities, and maintain persistence, reducing the need for continuous human control.
  3. Prompt injection will be treated as a practical attack technique against AI deployments. As organisations embed AI assistants and agents into workflows, attackers are expected to target the AI layer itself (e.g. through prompt injection, unsafe tool use, and weak guardrails) to trigger unintended actions or expose data.
  4. Threat actors will use AI to target humans at scale. The text emphasises AI-enhanced social engineering: conversational bots, real-time manipulation, and automated account takeover, shifting attacks from isolated human-led attempts to continuous, scalable interaction.
  5. AI will expose APIs as a too-easily-exploited attack surface. The experts argue that AI agents capable of discovering and interacting with software interfaces will lower the barrier to abusing APIs, including undocumented or unintended ones. As agents gain broader permissions and access to cloud services, APIs are expected to become a more frequent point of exploitation and concealment.
  6. Extortion will evolve beyond ransomware encryption. Extortion campaigns are expected to rely less on encryption alone and more on a combination of tactics, including data theft, threats to leak or alter information, and disruption of cloud services, backups, and supply chains.
  7. Cyber incidents will increasingly spread from IT into industrial operations. Ransomware and related intrusions are expected to move beyond enterprise IT systems and disrupt operational technology and industrial control environments, amplifying downtime, supply-chain disruption, and operational impact.
  8. The insider threat will increasingly include imposter employees. Analysts anticipate insider risks will extend beyond malicious or negligent staff to include external actors who gain physical or remote access by posing as legitimate employees, including through hardware implants or direct device access that bypasses end point security.
  9. Nation-state cyber activity will continue to target Western governments and industries. Experts point to continued cyber operations by state-linked actors, including financially motivated campaigns and influence operations, with increased use of social engineering, deception techniques, and AI-enabled tools to scale and refine targeting.
  10. Identity management is expected to remain a primary failure point. The rapid growth of human and machine identities, including AI agents, across SaaS, cloud platforms and third-party environments is likely to reinforce credential misuse as a leading cause of major breaches.

Taken together, these trends suggest that in 2026, cyber risk will increasingly reflect systemic exposure created by the combination of AI adoption, identity sprawl, and interconnected digital infrastructure, rather than isolated technical failures.

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

EU cyber rules target global tech dependence

The European Union has proposed new cybersecurity rules aimed at reducing reliance on high-risk technology suppliers, particularly from China. In the European Union, policymakers argue existing voluntary measures failed to curb dependence on vendors such as Huawei and ZTE.

The proposal would introduce binding obligations for telecom operators across the European Union to phase out Chinese equipment. At the same time, officials have warned that reliance on US cloud and satellite services also poses security risks for Europe.

Despite increased funding and expanded certification plans, divisions remain within the European Union. Countries including Germany and France support stricter sovereignty rules, while others favour continued partnerships with US technology firms.

Analysts say the lack of consensus in the European Union could weaken the impact of the reforms. Without clear enforcement and investment in European alternatives, Europe may struggle to reduce dependence on both China and the US.

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

New EU cybersecurity package strengthens resilience and ENISA powers

The European Commission has unveiled a broad cybersecurity package that moves the EU beyond certification reform towards systemic resilience across critical digital infrastructure.

Building on plans to expand EU cybersecurity certification beyond products and services, the revised Cybersecurity Act introduces a risk-based framework for securing ICT supply chains, with particular focus on dependencies, foreign interference, and high-risk third-country suppliers.

A central shift concerns supply-chain security as a geopolitical issue. The proposal enables mandatory derisking of mobile telecommunications networks, reinforcing earlier efforts under the 5G security toolbox.

Certification reform continues through a redesigned European Cybersecurity Certification Framework, promising clearer governance, faster scheme development, and voluntary certification that can cover organisational cyber posture alongside technical compliance.

The package also tackles regulatory complexity. Targeted amendments to the NIS2 Directive aim to ease compliance for tens of thousands of companies by clarifying jurisdictional rules, introducing a new ‘small mid-cap’ category, and streamlining incident reporting through a single EU entry point.

Enhanced ransomware data collection and cross-border supervision are intended to reduce fragmentation while strengthening enforcement consistency.

ENISA’s role is further expanded from coordination towards operational support. The agency would issue early threat alerts, assist in ransomware recovery with national authorities and Europol, and develop EU-wide vulnerability management and skills attestation schemes.

Together, the measures signal a shift from fragmented safeguards towards a more integrated model of European cyber sovereignty.

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

Researchers report increased ransomware and hacktivist activities targeting industrial systems in 2025

Industrial technology environments experienced a higher volume of cyber incidents in 2025, alongside a reported doubling in the exploitation of industrial control system (ICS) vulnerabilities.

According to the Cyble Research & Intelligence Labs Annual Threat Landscape Report 2025, manufacturing and healthcare (both highly dependent on ICS) were the sectors most affected by ransomware. The report recorded a 37% increase in total ransomware incidents between 2024 and 2025.

The analysis shows that the increase in reported ICS vulnerabilities is partly linked to greater exploitation by threat actors targeting human–machine interfaces (HMIs) and supervisory control and data acquisition (SCADA) systems. Over the reporting period, 600 manufacturing entities and 477 healthcare organizations were affected by ransomware incidents.

In parallel, hacktivist activity targeting ICT- and OT-reliant sectors, including energy, utilities, and transportation, increased in 2025. Several groups focused on ICS environments, primarily by exposing internet-accessible HMIs and other operational interfaces. Cyble further noted that 27 of the disclosed ICT vulnerabilities involved internet-exposed assets across multiple critical infrastructure sectors.

The report assessed hacktivism as increasingly coordinated across borders, with activity patterns aligning with geopolitical developments. Cyber operations linked to tensions between Israel and Iran involved 74 hacktivist groups, while India–Pakistan tensions were associated with approximately 1.5 million intrusion attempts.

Based on these observations, Cyble researchers assess that in 2026, threat actors are likely to continue focusing on exposed HMI and SCADA systems, including through virtual network computing (VNC) access, where such systems remain reachable from the internet.

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

New ETSI standard defines cybersecurity rules for AI systems

ETSI has released ETSI EN 304 223, a new European Standard establishing baseline cybersecurity requirements for AI systems.

Approved by national standards bodies, the framework becomes the first globally applicable EN focused specifically on securing AI, extending its relevance beyond European markets.

The standard recognises that AI introduces security risks not found in traditional software. Threats such as data poisoning, indirect prompt injection and vulnerabilities linked to complex data management demand tailored defences instead of conventional approaches alone.

ETSI EN 304 223 combines established cybersecurity practices with targeted measures designed for the distinctive characteristics of AI models and systems.

Adopting a full lifecycle perspective, the ETSI framework defines thirteen principles across secure design, development, deployment, maintenance and end of life.

Alignment with internationally recognised AI lifecycle models supports interoperability and consistent implementation across existing regulatory and technical ecosystems.

ETSI EN 304 223 is intended for organisations across the AI supply chain, including vendors, integrators and operators, and covers systems based on deep neural networks, including generative AI.

Further guidance is expected through ETSI TR 104 159, which will focus on generative AI risks such as deepfakes, misinformation, confidentiality concerns and intellectual property protection.

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

Western cyber agencies issue guidance on cyber risks to industrial sectors

A group of international cybersecurity agencies has released new technical guidance addressing the security of operational technology (OT) used in industrial and critical infrastructure environments.

The guidance, led by the UK’s National Cyber Security Centre (NCSC), provides recommendations for securely connecting industrial control systems, sensors, and other operational equipment that support essential services. The NCSC operates as part of GCHQ, one of the three UK Intelligence and Security Agencies.

Operational technology is widely used across sectors such as energy generation, water treatment, manufacturing, and transportation. While many of these systems were previously isolated from external networks, they are increasingly connected for remote monitoring, maintenance, and performance optimisation. The shift has expanded the number of network interfaces that require protection.

According to the co-authoring agencies, industrial environments are being targeted by a range of actors, including cybercriminal groups and state-linked actors. The guidance references a joint advisory issued in June 2023 on China-linked cyber activity, as well as a more recent advisory from the US Cybersecurity and Infrastructure Security Agency (CISA) that notes opportunistic activity by pro-Russia hacktivist groups affecting critical infrastructure globally.

The document was co-authored by agencies from multiple countries, including the United States, the Netherlands, Germany, and members of the Five Eyes partnership, Australia, Canada, and New Zealand, alongside UK authorities.

In an accompanying statement, NCSC Chief Technology Officer Ollie Whitehouse noted that cybersecurity measures in industrial environments should be integrated into system design and operations to support safety, availability, and service continuity.

The agencies state that the guidance is intended for OT operators, system integrators, and security professionals responsible for designing or maintaining connected industrial systems, and is applicable across sectors and jurisdictions.

The guidance outlines eight technical principles to reduce risk in connected OT environments. These include network segmentation, strong authentication mechanisms, continuous monitoring, and limiting remote access pathways. The agencies note that these measures are intended to reduce the likelihood of disruptive incidents affecting essential services.

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

One-click vulnerability in Telegram bypasses VPN and proxy protection

A newly identified vulnerability in Telegram’s mobile apps allows attackers to reveal users’ real IP addresses with a single click. The flaw, known as a ‘one-click IP leak’, can expose location and network details even when VPNs or proxies are enabled.

The issue comes from Telegram’s automatic proxy testing process. When a user clicks a disguised proxy link, the app initiates a direct connection request that bypasses all privacy protections and reveals the device’s real IP address.

Cybersecurity researcher @0x6rss demonstrated an attack on X, showing that a single click is enough to log a victim’s real IP address. The request behaves similarly to known Windows NTLM leaks, where background authentication attempts expose identifying information without explicit user consent.

Attackers can embed malicious proxy links in chats or channels, masking them as standard usernames. Once clicked, Telegram silently runs the proxy test, bypasses VPN or SOCKS5 protections, and sends the device’s real IP address to the attacker’s server, enabling tracking, surveillance, or doxxing.

Both Android and iOS versions are affected, putting millions of privacy-focused users at risk. Researchers recommend avoiding unknown links, turning off automatic proxy detection where possible, and using firewall tools to block outbound proxy tests. Telegram has not publicly confirmed a fix.

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

Cyber Fortress strengthens European cyber resilience

Luxembourg has hosted its largest national cyber defence exercise, Cyber Fortress, bringing together military and civilian specialists to practise responding to real-time cyberattacks on digital systems.

Since its launch in 2021, Cyber Fortress has evolved beyond a purely technical drill. The exercise now includes a realistic fictional scenario supported by media injections, creating a more immersive and practical training environment for participants.

This year’s edition expanded its international reach, with teams joining from Belgium, Latvia, Malta and the EU Cyber Rapid Response Teams. Around 100 participants also took part from a parallel site in Latvia, working alongside Luxembourg-based teams.

The exercise focuses on interoperability during cyber crises. Participants respond to multiple simulated attacks while protecting critical services, including systems linked to drone operations and other sensitive infrastructure.

Cyber Fortress now covers technical, procedural and management aspects of cyber defence. A new emphasis on disinformation, deepfakes and fake news reflects the growing importance of information warfare.

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