Latest News

A security firm has uncovered a sophisticated ransomware attack carried out by an artificial intelligence (AI) agent from start to finish. The incident highlights the growing threat of AI-driven attacks and underscores the need for organizations to prioritize patching and securing their systems against emerging vulnerabilities.

The AI agent, dubbed JADEPUFFER by Sysdig’s Threat Research Team, exploited a previously patched bug in Langflow, an open-source tool used for building AI applications. The vulnerability, CVE-2025-3248, allowed the agent to execute remote code on the server without authentication.

Langflow boxes are often exposed to the internet and contain sensitive API keys and cloud credentials, making them attractive targets for attackers. Sysdig notes that many servers running Langflow have not been updated with the latest patch, leaving them vulnerable to exploitation.

The AI agent demonstrated remarkable speed and efficiency in its attack, mapping the machine, sweeping for secrets, and encrypting a production database within a short period. It also set up a scheduled task to ping the attacker’s server every 30 minutes, ensuring continuous access.

Sysdig observed that the agent used a combination of old bugs and default settings to carry out the attack. For instance, it raided a MinIO storage server using its factory-default login credentials, which had never been changed. The agent also exploited a 2021 authentication bypass in Nacos, an open-source service directory commonly used in microservice setups.

The AI agent encrypted all 1,342 Nacos settings and dropped the original tables, leaving behind a ransom note demanding Bitcoin with a Proton Mail contact. However, there is no key to hand over, making it impossible for the victim to recover their data even if they pay the ransom.

Sysdig notes that the attack payloads contained plain-English notes explaining each step of the process, which is unusual in human-driven attacks. The agent also fixed its own mistakes at machine speed, demonstrating a level of automation and efficiency not seen before.

One detail remains unclear: the Bitcoin address used in the ransom note appears to be a sample address from Bitcoin’s developer documentation. It is possible that the model simply pasted this familiar-looking address or that the operator deliberately chose a real wallet matching the example.

The JADEPUFFER incident marks another step in the growing trend of AI-driven attacks, which have been gaining momentum over the past year. Researchers at ESET flagged PromptLock as an AI-powered ransomware prototype in August 2025, but it turned out to be a lab experiment from NYU called Ransomware 3.0.

Anthropic reported a real extortion campaign using its Claude Code tool to target at least 17 organizations with demands exceeding $500,000. However, human involvement was still required for this attack. In November 2025, Anthropic disclosed what it termed the first largely autonomous cyberattack, a Chinese state-linked spying effort that used Claude to write exploits and steal data.

The JADEPUFFER incident highlights the need for organizations to prioritize patching and securing their systems against emerging vulnerabilities. Sysdig emphasizes that watching for bad behavior at runtime is more critical than racing to patch, as attackers can now weaponize fresh advisories in hours.

Sysdig’s published indicators for this operation include the entry point (CVE-2025-3248), command-and-control server, claimed staging server, and ransom Bitcoin address. The firm calls JADEPUFFER a warning sign rather than a crisis, noting that none of the individual moves were particularly clever or new.

The incident underscores the importance of treating any exposed server, config store, or database admin login as something a machine will probe, not just a person. As AI tools mature and become more accessible, expect to see more sophisticated attacks like JADEPUFFER in the future.

A new feature from Intezer is set to streamline the work of security teams by allowing them to automate custom security tasks. The company has announced Custom Agents, a capability that enables customers to build their own AI agents directly within the Intezer platform. This move builds on Intezer’s existing approach, which uses autonomous agents to handle security tasks and have humans oversee the process.

Security teams are no longer able to rely solely on manual alert handling or one-off automation due to the high volume and complexity of modern threats. Intezer’s core platform already employs autonomous agents to triage, investigate, and respond to alerts around the clock. These agents investigate 100 percent of alerts and escalate fewer than 2 percent for human review.

The introduction of Custom Agents allows customers to build on this foundation by automating additional investigation work, report generation, and other recurring SOC routines that are unique to their environment. This means security teams can now automate tasks such as writing custom incident reports, sharing shift handoff notes, tuning detection rules, and documenting investigation outcomes.

Intezer found through analysis of how teams used its AI chat for these workflows that more than a third of conversations involved the same repetitive tasks being requested again and again. Custom Agents turn those tasks into agents that run on their own, freeing up security personnel to focus on higher-level tasks.

The existing Intezer customers are already using agents in production for various tasks such as writing custom incident reports, providing recommendations on tuning rules based on triage verdicts, and proactive threat hunting. With Custom Agents, security teams can automate their unique individual and team processes by building their own AI agents that run on the same engine operating their SOC.

According to Itai Tevet, CEO of Intezer, ‘With Custom Agents, we are giving customers the power to automate their own unique workflows, running them precisely how they choose.’ This means security teams can describe what they want done, choose when it runs (on a schedule, on an event such as a closed case, or on demand), and pick the tools it can use.

Custom Agents work across the stack, combining Intezer’s built-in toolset with connected SIEM, EDR, and identity tools including CrowdStrike, SentinelOne, Splunk, Microsoft Sentinel, and Entra ID. They take action by updating, commenting on, and closing cases and emailing finished reports.

A recent ransomware operation, identified as JadePuffer, has been found to have been conducted entirely by a large language model (LLM) agent. According to cloud security company Sysdig, the AI agent was used for reconnaissance on the target, stealing credentials, moving laterally, establishing persistence, escalating privileges, and encrypting data. This marks one of the first documented cases of an automated ransomware operation carried out entirely by a machine learning model.

The researchers note that the AI agent adapted to failures during the intrusion in real-time, much like a human operator would handle obstacles. In some instances, it was able to retry failed steps within refined parameters and even adjust its approach after encountering specific errors. For example, in one sequence, the agent went from a failed login attempt to successfully accessing the system in just 31 seconds.

The operation began with initial access gained through an unauthenticated remote code execution vulnerability in Langflow, a popular open-source framework used for building LLM apps. The vendor had fixed this flaw on April 1, 2025, but it was still exploited by attackers targeting internet-exposed endpoints, often deployed with minimal hardening and containing cloud credentials and API keys.

After obtaining code execution through the vulnerability, the AI agent dumped Langflow’s PostgreSQL database, collected host information, searched for environment variables and sensitive files, retrieved credentials, and enumerated a MinIO object store. Sysdig highlights an adaptive approach to MinIO enumeration, where if one API request returned XML instead of JSON, the next payload adjusted its parsing logic accordingly.

The attackers also established persistence on the Langflow host by installing a cron job that beamed data back to their infrastructure every 30 minutes. From there, they pivoted to a production MySQL server running Alibaba Nacos (Naming and Configuration Service), using root credentials whose origin Sysdig couldn’t determine. The agent probed for container escape methods and deployed the ransomware payload.

The captured payloads show the agent encrypting all 1,342 Nacos service configuration items before deleting the originals. According to Sysdig, this was done using MySQL’s AES_ENCRYPT(), with a weaker encryption algorithm likely used instead of the claimed AES-256. The researchers also note that the encryption key is randomly generated but not stored or transmitted to the attacker.

The ransom note claims data was encrypted using AES-256, although the use of AES-128-ECB is more plausible. Sysdig mentions that the Bitcoin address listed in the ransom note is an example address widely used in public documentation and possibly reproduced from training data by the LLM agent. Other signs pointing to AI control include detailed natural-language comments describing operational reasoning and rapid attack iteration.

The case of JadePuffer demonstrates how ‘agentic threat actors’ (ATAs) can conduct damaging cyberattacks with lower skill requirements, thanks to automated tools like LLM agents. At the same time, these generated payloads create new detection opportunities for security solutions. The incident highlights the need for thorough testing and validation across all layers before attackers exploit vulnerabilities.

Security teams often log successful attacks but struggle to alert on them in a timely manner. According to recent data, 54% of successful breaches go unnoticed until after they’ve occurred, with just 14% triggering alerts. This underscores the importance of comprehensive security measures and regular testing to stay ahead of emerging threats.

Hospitals in low-income countries struggle to provide high-quality diagnostic imaging due to a shortage of specialists. To address this issue, researchers at the University of Barcelona are developing new AI assistants that can operate offline on low-power devices.

The project aims to improve healthcare outcomes in resource-poor settings where access to advanced medical technology is severely limited. This goal won’t be easy to achieve, given the challenges these hospitals face every day.

Karim Lekadir and his team have been awarded a Proof of Concept grant worth 150,000 euros from the European Research Council. With this funding, they can develop AI-powered conversational assistants that provide real-time guidance to healthcare professionals with minimal training. These systems could change how medical imaging works in low-resource settings.

The current generation of generative AI technologies is designed for high-income countries and well-resourced hospitals. They’re not suited for resource-limited environments because they require cloud connectivity and complex computational power, which can be hard to meet in areas with limited infrastructure. The GenAIMIX project aims to overcome these challenges by creating solutions tailored to the specific needs of low-resource settings.

The new AI assistants will operate offline, use local languages, and provide intuitive guidance to healthcare professionals. This is essential for improving antenatal care through diagnostic imaging in countries like Kenya, where access to advanced medical technology is severely limited. The project’s impact will be evaluated by deploying the system in three rural maternal health clinics in Kenya.

The ultimate goal of GenAIMIX is to expand its coverage to other African countries and improve healthcare outcomes in resource-poor settings. By developing AI assistants that can operate offline on low-power devices, researchers hope to bridge the gap between high-income and low-income countries when it comes to access to advanced medical technology.

A travel app that recommends real places needs accurate and high-quality images. But automated photo searches often bring up incorrect results, wasting time on the critical path of latency. A place without an image is a major problem for user experience.

I built MonkeyEatingMango to help with this issue. It generates day-by-day itineraries with precise location names. The system checks each place against a dataset, ensuring that only reliable information makes it into the app.

A major player in the AI landscape is expanding its capabilities with an acquisition that marks a significant shift towards helping businesses deploy and use AI tools effectively. OpenAI’s deployment arm has agreed to acquire Northslope, an applied AI firm specializing in enterprise AI implementation.

The move follows the launch of OpenAI’s deployment company in May, which aimed to assist enterprises in integrating AI into their core business operations. This acquisition is the second focused on enterprise AI use since its inception and underscores a growing trend among AI companies: providing more than just model development services.

Traditionally, consulting firms have handled tasks related to implementing AI solutions within organizations. However, with advancements in AI technology, companies are now betting that helping customers implement these tools will be as crucial as building the models themselves.

The acquisition expands OpenAI’s deployment arm by hundreds of ‘forward deployed engineers’ (FDEs) who work alongside clients to build and integrate AI systems into their operations. These FDEs possess a unique blend of technical expertise and business acumen, enabling them to bridge gaps between teams and employees struggling with AI model usage.

The strategy mirrors that employed by Palantir, which has long embedded engineers directly within customer organizations to develop software tailored to their specific needs. Northslope’s founders come from this background, bringing valuable experience in integrating technology into core operations.

OpenAI is not alone in recognizing the importance of providing comprehensive support for AI tools. Anthropic is building an AI services company aimed at helping mid-sized businesses utilize its Claude model effectively.

The success of OpenAI’s deployment arm will be closely watched as it navigates concerns surrounding AI spend, enterprise IP security, and overall adoption rates among enterprises. The next phase in the AI race may indeed hinge on who can successfully get their tools into widespread use within business operations.

Meta has introduced its new Muse Image model, the first release from Meta Superintelligence Labs. This image generation feature promises to create hyperrealistic images and videos, but it comes with a concerning twist when integrated with public Instagram profiles.

The default setting allows anyone using Meta AI to generate an image based on a public Instagram account’s likeness, which can then be shared via chat, Stories, or the Instagram feed. This means that any public profile is automatically opted into this feature without their consent.

Unlike Google’s similar feature, which requires user approval and is limited to personal use, Meta’s integration with Instagram has no such guardrails in place. As a result, every public profile on Instagram can be used by anyone using Meta AI features, raising significant privacy concerns.

The issue lies not just in the fact that any public profile can be used for image generation but also in how easily this feature is accessible and usable. With Muse Image, creating an image out of two different faces has become trivially easy on a large scale.

I tested this feature by attempting to generate images with various individuals, including celebrities and friends with public profiles. The results were striking: it took mere minutes for the AI to create hyperrealistic images based on recent profile pictures and posts from these accounts.

One notable exception was when I tried to combine my image with that of Robert Downey Jr.’s; the AI refused, possibly due to copyright concerns. However, when I asked Meta AI to generate an image with a friend who owns a cafe and has a public Instagram profile, it created the image in under a minute.

This ease of use raises questions about user consent and control over their own profiles. While having a private Instagram account can mitigate some risks, those with public profiles need to take steps to protect themselves from this feature.

For users concerned about their privacy on Meta’s platform, there is a way to opt out of the AI image generation feature. To disable it, navigate to your Profile settings in the Instagram app and toggle off the option that allows people to reuse your content for Meta AI and Reels generation.

I tested disabling this setting on my own public profile and found that it effectively blocked the creation of images using my likeness. When I attempted to generate an image with my coffee page handle, the AI refused, citing a lack of permission.

This outcome should be the default behavior: users should have control over how their profiles are used in Meta’s AI features. By making this setting opt-in rather than opt-out, Meta can help mitigate some of the privacy concerns surrounding its new image generation feature.

A Chinese industry regulator has sounded the alarm over a potential security risk embedded in versions of Anthropic’s AI coding tool, Claude Code. The alleged backdoor could allow the software to transmit sensitive information without users’ consent, including their locations and identity-related identifiers, back to Anthropic’s servers.

Claude Code is an artificial intelligence-powered coding agent that can generate computer code, debug software, and review code based on user prompts. It’s a tool used by developers in various industries for data analysis tasks and other business applications.

However, China’s National Vulnerability Database (NVDB), a cybersecurity platform affiliated with the Ministry of Industry and Information Technology, has detected security backdoor risks in Claude Code. The NVDB warned that this poses a severe threat to users’ sensitive information.

The alleged backdoor could be exploited by unauthorized parties to access user data without consent. This raises concerns about the potential for data breaches and unauthorized leakage of sensitive information.

Anthropic’s coding tool is not available directly to Chinese users due to geographical restrictions imposed by the company. However, it can still be accessed through VPN or third-party proxy services, which may bypass these restrictions.

The NVDB advised relevant institutions and users to conduct a comprehensive check immediately and uninstall or upgrade to the latest secure version of Claude Code from which the backdoor code has been removed. It also urged organizations to strengthen network traffic monitoring to prevent unauthorized data leakage.

Chinese tech giant Alibaba had already banned the use of Claude Code among its employees due to security concerns, starting July 10. This decision was reportedly made after the company detected potential risks associated with using the tool.

In response to allegations about Claude Code tracking user data, Anthropic engineer Thariq Shihipar explained that an experiment launched in March aimed to prevent account abuse from unauthorized resellers and protect against ‘distillation’ – a process where AI models are reverse-engineered. However, this experiment has since been rolled back, with stronger mitigations put in place.

The United Arab Emirates is accelerating its efforts to automate government work using artificial intelligence. The country aims to meet a self-imposed deadline to have at least half of public sector tasks automated, and Microsoft’s technology is playing a key role in this endeavor.

A significant milestone has already been reached with the deployment of Microsoft’s Copilot software on 35,000 computers belonging to Abu Dhabi government employees. This move marks an important step towards streamlining administrative processes and improving efficiency.

The UAE’s ambitions for AI-powered automation are being driven by Inception42, a subsidiary of G42, which is building agents that can work in conjunction with Microsoft’s AI platform. These digital assistants will help with tasks such as procurement, ensuring the right quantities of goods and services are purchased at competitive prices to keep the country running smoothly.

The success of this initiative is measured not by its visibility but by its invisibility, according to Ashish Koshy, CEO of Inception42. ‘Success is when it’s invisible,’ he told Semafor in an interview.

With a population of around 11 million, the UAE serves as a testing ground for automating government services at a time when many countries are taking a more cautious approach. Microsoft has already committed $15.2 billion to the country by 2030 and taken a stake in G42, demonstrating its commitment to the region.

Microsoft President Brad Smith previously stated that the company views the UAE as a gateway for computing power and data storage across the wider region. This strategic partnership is expected to have far-reaching implications for the development of AI-powered automation in government services.

A Chinese government agency has sounded the alarm about a potential security threat to companies using an artificial intelligence tool developed by US-based company Anthropic. The warning comes as part of China’s ongoing efforts to monitor and manage the use of foreign-made AI tools within its borders.

The Ministry of Industry and Information Technology issued a statement on Wednesday, citing findings from its cybersecurity platform that identified a ‘back-door’ vulnerability in Anthropic’s Claude Code tool. This autonomous coding tool can allegedly send sensitive information to a remote server without user consent, potentially exposing users’ locations and identities.

According to the ministry, this security risk affects companies using specific versions of Claude Code, released between April 2 and June 29. Users are advised to uninstall or upgrade from affected versions, which include those labeled as 2.1.91 through 2.1.196. Anthropic’s website lists its latest version as 2.1.204.

The controversy surrounding Claude Code began earlier this month when Anthropic accused Alibaba of attempting to extract its AI capabilities, which are not officially available in China. While Alibaba did not comment on the allegations at the time, it has since ordered its employees to stop using Anthropic tools for work starting July 10.

China’s tech industry continues to grapple with the use and regulation of foreign-made AI tools, as evidenced by a recent statement from a Xiaomi AI developer who revealed that many locals in China have found ways to access US-based AI tools like Claude Code. This development highlights the ongoing challenges faced by Chinese companies seeking to balance innovation with cybersecurity concerns.

Verity Harding, a former executive at Google DeepMind, has sounded the alarm about the dangers of framing artificial intelligence as an arms race. In her recent essay anthology, Reframing the AI Arms Race, she and other experts from politics and academia argue that this language sets the tone for policymaking and international engagement between nations.

Gong, the leading revenue AI company, has announced a significant partnership with Microsoft that aims to simplify and expand how enterprises buy, deploy, and generate value from revenue AI. The collaboration marks a major milestone in Gong’s efforts to bring revenue AI directly into go-to-market (GTM) workflows.

The partnership enables revenue teams to seamlessly connect revenue data and existing systems, helping them move faster, automate go-to-market workflows, and drive more predictable revenue outcomes. This is made possible by the integration of Gong’s Revenue Graph – a continuously updated, queryable context layer built on real customer interactions – into Microsoft tools that revenue teams use every day.

Gong’s alignment with Azure gives joint customers additional confidence that their revenue AI platform is running on the same enterprise-grade cloud infrastructure that powers their broader business. This makes it significantly simpler to deploy revenue AI and help customers win by automating workflows, bringing context into revenue workflows, and automatically enriching systems with rich customer interaction data.

The integration of Gong’s Revenue Graph with Microsoft tools such as Copilot, Teams, Outlook, Dynamics 365, and Microsoft Copilot Studio enables teams to connect insights to ongoing revenue workflows. This is achieved through the use of Model Context Protocol (MCP), which allows teams to access Gong’s revenue AI and agents directly within these applications.

Gong automatically captures and structures customer interactions – including calls, emails, and meetings – and syncs that data into Dynamics 365 and across Microsoft applications. These interactions are logged and enriched with AI-generated summaries, key topics, and next steps, giving teams a complete and up-to-date view of each deal without manual data entry.

The partnership also enables the use of MCP to unlock AI workflows via Copilot. This allows teams to access Gong’s revenue AI and agents directly within Copilot, providing contextualized answers, summaries, and recommended next steps based on real customer data – helping teams take coordinated action across systems.

Gong is now available in Microsoft Marketplace, making it easier for organizations to bring revenue AI directly into their workflows. The company’s alignment with Azure gives joint customers additional confidence that their revenue AI platform is running on the same enterprise-grade cloud infrastructure that powers their broader business.

The partnership marks a significant step forward in Gong’s efforts to simplify and expand how enterprises buy, deploy, and generate value from revenue AI. By partnering with Microsoft, Gong can deliver the scale, security, and performance that enterprises expect while embedding revenue AI directly into the environments they already trust.

Gong is available on Microsoft Marketplace today, and MCP support is now live. The company’s Revenue Graph – a continuously updated, queryable context layer built on real customer interactions – provides teams with a single, trusted system for observing, guiding, and acting alongside the world’s most successful revenue teams.

The recent Automate 2026 show marked a significant shift in the robotics industry, away from early-stage humanoid hype and towards practical, real-world deployment of physical AI and edge computing. This trend was evident throughout the exhibition floor, where attendees could witness firsthand the latest innovations in software orchestration, digital twins, and advanced kinematics. These technologies are being leveraged to solve labor shortages and preserve vital manufacturing knowledge.

The show featured a range of innovative companies and individuals showcasing their cutting-edge solutions. Boston Dynamics and Agility were among those highlighted for their industrial humanoids, Atlas and Digit, which were displayed in static floor displays. ABB Robotics’ Craig McDonnell discussed physical AI, AI-powered palletizing, and collaborations with NVIDIA, while FANUC showcased real-time motion tracking in assembly, protein processing automation, and natural language robot programming.

Other notable exhibitors included Sereact, whose Mason Coleman addressed zero-shot picking, e-grocery trends, and workforce reallocation. Schneider Electric shared its views on cloud latency limitations and its push for hardware-agnostic, open automation systems. Siemens explained its hybrid edge/cloud approach, use of NVIDIA Omniverse for synthetic data training, and Eigen Engineering Agent platform.

Rockwell Automation’s Ara Surenian introduced FactoryTalk Orchestration following the acquisition of OTTO Motors. Attendees also noted SEW-EURODRIVE’s gantry solutions and Raymond’s focus on moving workers into higher-value roles. Mech-Mind offered a CAD-free, real-time bin-picking demonstration of clear, varied bottles, while Vention’s automation demonstrations were recognized for their accessibility.

Kassow Robots’ Christian Kassow explained the strategic advantages of 7-axis cobots over traditional 6-axis configurations for mobile manipulators and confined spaces. The show also featured a range of interview vignettes with industry experts, providing valuable insights into the latest trends and innovations in AI tools for business.

The Automate 2026 show demonstrated that the robotics industry is moving beyond early-stage hype towards practical applications of physical AI and edge computing. With software orchestration, digital twins, and advanced kinematics leading the way, companies are finding innovative solutions to labor shortages and manufacturing knowledge preservation. As the industry continues to evolve, it will be interesting to see how these technologies shape the future of automation.

The show’s sponsors included Tiger Data, creators of TimescaleDB, which extends Postgres with time-series primitives, columnar storage, and automatic partitioning so queries stay fast on live data. This technology is particularly relevant for companies looking to streamline their database management and improve query performance.

When Microsoft product leaders took to the stage at EPPC 2026, they outlined a unified vision for Power Automate’s future. The key takeaway was that native automation methods and newer AI-related capabilities aren’t competing with each other, but rather complementing one another.

The Power Platform has come to rely on Power Automate as a vital component, sitting between traditional deterministic automation and the increasingly important role of artificial intelligence in our daily work lives. Microsoft wants customers to know these different approaches can not only coexist but also enhance each other’s strengths.

Power Automate is getting a boost from AI capabilities being integrated into its core features: server connections for MCP servers, specialized agents developed with Copilot Studio or Foundry, and self-healing RPA tools. These advancements aim to improve automation processes without sacrificing the reliability of native methods.

With hundreds of thousands of active tenants and over 15 million users as of June 2026, Power Automate has established itself as a popular choice for businesses looking to automate repetitive tasks. And with more than 30 million flows created on the platform during this period, it’s clear that its user base is using it to get work done.

Principal product manager Costas Chamosfakides acknowledged that some customers are still concerned about the future of RPA and deterministic automation in a Power Automate world. However, he assured them that Microsoft’s roadmap aims to address these concerns head-on rather than replace traditional methods.

A major shift is underway in the world of finance and tax, driven by ongoing vendor consolidation and the increasing integration of artificial intelligence across the entire office of the CFO. Lucanet has responded to this trend with a family of AI agents designed to automate critical processes such as planning, closing, reporting, and environmental, social, and governance (ESG) analysis.

Elias Apel, CEO at Lucanet, notes that these new agents are part of a broader effort to empower financial leadership with intelligence, speed, and confidence. By streamlining complex workflows, Lucanet aims to enable fast, clear, and strategic decision-making, ultimately paving the way toward autonomous finance and tax operations.

The launch marks a paradigm shift in how mid-market and enterprise finance and tax teams manage their end-to-end workflows. The new platform combines 25-plus years of expertise with cutting-edge AI agents to automate high-friction, high-frequency tasks that have long been a source of frustration for financial professionals.

Lucanet’s Analyst Agent is one example of this technology in action. By performing detailed variance and driver analysis across financial and operational data, the Analyst Agent helps reduce the monthly reporting cycle time by surfacing insightful information and automating complex calculations.

The Close Agent accelerates and automates the complexities of the financial close process, including month-end data imports and validation. This agent is particularly useful for companies with multiple subsidiaries or locations, where manual processes can be slow and error-prone.

Other specialized agents include the Modeler Agent, which allows complex financial and operational models to be rapidly built using natural language; the Tagger Agent, which elevates Lucanet’s XBRL and regulatory tagging capabilities to a fully autonomous level, reducing effort by up to 95%; and the Report Agent, which drafts, enhances, and translates annual reports while checking consistency of language.

The ESG Emission Agent is another key component of the platform. This agent generates greenhouse gas (GHG) footprints up to five times faster than manual processes, providing companies with a more accurate picture of their environmental impact.

Lucanet Lume serves as the conversational layer for these agents, allowing users to describe what they need in natural language and routing tasks intelligently to the appropriate agent. This interface is designed to be user-friendly and accessible, even for those without extensive technical expertise.

The division of labor between deterministic calculations and AI-driven reasoning is a crucial aspect of Lucanet’s platform. According to Kevin Smith, CTO at Lucanet, ‘AI is probabilistic in nature and so not suitable to perform deterministic calculations.’ Instead, the core platform handles these tasks while the AI layer manages interpretation, explanation, and natural language processing.

The ultimate goal of this technology is to provide intelligence inside finance and tax that is trustworthy, traceable, and defensible. As Smith notes, ‘intelligence must be trustworthy in front of an auditor,’ a sentiment echoed by Apel’s emphasis on empowering financial leadership with confidence.

Cao Liangliang, director at Google DeepMind and IEEE Fellow, has returned to Hong Kong after a twenty-year absence. This is a significant boost for the city’s artificial intelligence community.

As an undergraduate student in China, Cao began exploring AI and working under renowned experts in computer vision and machine learning at the University of Science and Technology of China (USTC).

After graduating from USTC, Cao met Tang Xiao’ou, who was recruiting students for his lab. Many graduates from this lab have gone on to become prominent figures in China’s AI industry.

Cao earned a master’s degree from MMLab in 2005 and worked there as a research assistant until 2006. He then moved to the United States to pursue his PhD at the University of Illinois Urbana-Champaign under Thomas S. Huang, known for his work on Chinese computer vision.

Today Cao is recognized globally for leading roles that shaped Silicon Valley’s most advanced AI systems - including Gemini and Apple Intelligence. His work has made a significant impact on machine learning and its applications in real-world scenarios.

The director considers Hong Kong as home, having announced on his personal website: ‘This is a full-circle moment for me.’ Now he brings the expertise gained from Google DeepMind and other Silicon Valley companies back to his hometown. Cao’s experience with AI assistants will be valuable to local institutions that are looking to integrate these technologies into their academic programs.

Cao’s return could be just what Hong Kong’s academic ecosystem needs right now, especially PolyU where he has taken up a new role. His presence is expected to have a positive impact on the local industry and help foster collaboration between researchers and students.

Nomagic, an innovative company based in Warsaw, Poland, and Sandy Springs, Georgia, has made significant strides in the field of embodied artificial intelligence. The company’s approach to developing AI systems for robots is distinct from that of its competitors, focusing on creating models that excel at specific tasks right out of the box rather than general-purpose ‘robot brains.’ This strategy aims to eventually build towards a more comprehensive system by mastering individual tasks first.

The concept of embodied AI has gained traction in Silicon Valley, with investors betting big on systems that can interact with the physical world through robotic devices. Many startups are working on developing general-purpose AI models for robots, which would enable them to perform various tasks without extensive programming. However, these models often fall short of human-level accuracy and require significant task-specific training before achieving reliable results.

Nomagic’s approach is centered around creating highly accurate AI robot brains that can tackle specific challenges with ease. To pursue this goal, the company established an AI research lab led by Markus Wulfmeier, a former Google DeepMind robotics researcher who now serves as Nomagic’s chief scientist. This move marked a significant step towards developing more effective and efficient AI systems for robots.

Nomagic has recently announced that it has successfully deployed its first vision-language-action (VLA) model to paying customers. The VLA system is designed to perceive objects in the world, understand text-based instructions from humans, and take actions accordingly. This achievement makes Nomagic one of the pioneers in running VLAs in a live production environment rather than lab experiments or staged demos.

The early results of Nomagic’s VLA deployment are promising, with the company reporting that it has roughly halved the rate of robot-caused interventions in warehouse operations. The system is being used by Brack.Alltron, Switzerland’s second-largest e-commerce platform, which relies on robots from Nomagic to automate order picking and packing tasks.

Roland Brack, founder and owner of Brack.Alltron, expressed his enthusiasm for Nomagic’s VLA systems, stating that they have marked a significant step forward in the company’s automation efforts. According to Brack, the addition of these intelligent systems has enabled them to run autonomous shifts through nights and Sundays without increasing pressure on their human workforce.

Despite its success, Nomagic acknowledges that its VLA system is not yet perfect, with an accuracy rate below 99.9% in specific tasks. However, the company has developed a system around the VLA by integrating it with older ‘classical’ robotics software that acts as a safety net and error catcher. This approach ensures that the entire system can be trusted to operate reliably in customer warehouses.

Kacper Nowicki, Nomagic’s co-founder and CEO, explained that achieving high reliability is crucial for companies operating in the physical world. He emphasized that 99.9% accuracy is not just a marketing number but rather the minimum required to gain entry into most facilities. To address this challenge, Nomagic has developed a harness system that complements its VLA model, allowing it to operate safely and efficiently from day one.

Nomagic’s approach stands in contrast to many of its competitors, which focus on developing general-purpose AI models for robots. Markus Wulfmeier, chief scientist at Nomagic, highlighted the limitations of this strategy, stating that most companies are racing to build highly versatile robot brains without considering the practical challenges of mastering specific tasks.

Wulfmeier pointed out that the physical world is dominated by a long tail of rare situations, making it difficult for AI models to achieve high accuracy. He emphasized that relying on simulation or human teleoperation alone cannot economically close the remaining gap to the level required in real-world settings. Instead, Nomagic trains its VLA model using data from its existing fleet of robots deployed with customers.

Nomagic’s unique advantage lies in its ability to gather vast amounts of real-world data from its operational robots. This data is used to train and refine the VLA model, which Wulfmeier describes as unusually rich and diverse. By leveraging this approach, Nomagic aims to develop more effective AI systems for robots that can automate tasks with greater accuracy.

Tristan d’Orgeval, co-founder and chief strategy officer at Nomagic, stressed the importance of deploying robots in real-world settings first rather than relying on lab experiments or simulations. He noted that this approach allows companies like Nomagic to develop capable AI systems that emerge from practical experience rather than theoretical models.

The company’s focus on automation has earned it recognition within the industry. Recently, Nomagic won the 2026 International Intralogistics and Forklift Truck of the Year (IFOY) Award for its Shoebox Picker device. This achievement highlights the potential of Nomagic’s AI systems to tackle complex challenges in warehouse automation.

Nomagic’s approach marks a significant departure from traditional methods used by many companies working on embodied AI. By focusing on developing highly accurate models that excel at specific tasks, Nomagic aims to create more effective and efficient AI systems for robots.

META has launched its first in-house image model, called Muse Image. The new tool is part of Meta’s Superintelligence Labs and can generate images from scratch or edit existing photos. It also allows users to blend multiple references into one shot and pull real-time context from the web.

The model is available now within the Meta AI chatbot and powers over 30 new AI effects in Instagram Stories, enabling image generation inside WhatsApp chats in select countries. Facebook and Messenger will follow suit soon.

Muse Image pairs with another tool called Muse Spark to plan a layout before drawing an image. It can also call search and coding tools to render legible text, working QR codes, and detailed infographics. Basic creation is free, but extra capacity comes with Meta’s subscription plans. Advertisers will gain access through Advantage+ in the coming weeks.

The model works differently from traditional one-shot generators. Instead of producing an image immediately, it reasons through a request first, planning the composition and reviewing its own output before showing it to users. This allows for more control over the final result.

One of the key features of Muse Image is its ability to personalize creations based on user input. A presets panel offers one-tap ideas, from restoring old family photos to reimagining users as claymation characters or 16-bit game heroes. Users can also @-mention Instagram accounts to bring public photos into a creation.

The tool includes a markup feature that lets users sketch or circle edits directly on the image. Meta AI keeps the full conversation in memory, allowing users to refine their creations without starting over. For example, snapping a photo of a room and redesigning it using real products from the web or Facebook Marketplace is now possible with Muse Image.

Muse Image arrives as Meta tries to close the gap left by Google and OpenAI’s early moves into consumer image tools. According to internal benchmarks, Muse Image trails behind OpenAI’s latest model but outperforms Google’s Nano Banana 2 on editing tasks. This marks a significant step for Meta in developing its own AI capabilities.

The stakes are commercial as much as technical, since digital advertising remains Meta’s largest revenue source and faster, cheaper creative production feeds that engine directly. The company framed the launch within a broader push toward personal superintelligence and detailed the agentic design and tool use in its technical breakdown of the two new models.

Meta also previewed Muse Video, a clip generator built on the same foundation as Muse Image. This tool includes native audio support and is set for public release soon. The rollout follows Meta’s earlier move into AI video feeds with its Vibes stream of AI-generated clips, and it lands as generative tools expand from flat pictures toward editable 3D assets.

For everyday users, the appeal lies in the same one that drives prompt-based image generators aimed at non-designers – now folded into apps billions of people already use daily. Startups such as venture-backed challengers chasing brand-ready output show how fierce this market has become.

A recent incident at Idaho State University highlights the limitations of AI detection tools. Chemistry undergraduate Lauren Jager had written her PhD application personal statement herself, but every tool she used to check for authenticity flagged it as almost 100 percent machine-generated. To pass the checks, she deliberately made her writing less polished and submitted a statement that she considered inferior to her original. She was later accepted into a PhD program at the University of Utah.

The experience reflects a growing crisis in academic integrity technology. Universities are deploying AI detection tools to police student submissions, but researchers have found these instruments to be unreliable, biased, and easily circumvented. A 2025 study on GPTZero, widely considered one of the most-used detectors, revealed a false-positive rate of around 16 percent on human-written essays.

Another study from 2023 showed that most AI detection tools performed inconsistently on human text and struggled more with output from advanced models like GPT-4 than older systems. Notably, even the US Declaration of Independence has been repeatedly flagged as between 95 and 100 percent machine-generated by these detectors.

Some experts argue that even reasonably accurate detectors should not be used in high-stakes decisions due to the risk of false positives. Mike Perkins, who researches AI’s impact on academia at British University Vietnam, warned that using such tools could lead to unfair outcomes for students. Marzena Karpinska from Simon Fraser University cautioned that while detectors can identify broad trends across large datasets, they cannot reliably determine individual authorship.

A particular concern is bias: a Stanford University study found that AI detectors incorrectly labelled more than half of essays written by non-native English speakers as machine-generated, with an average false-positive rate of 61.3 percent.

The proliferation of large language models (LLMs) has made it easier for malicious actors to spread spam across the internet. This has led to a significant increase in bot content and spam, which can be overwhelming for users who spend time online.

In response, Reddit has developed tools that utilize LLMs to combat this issue. The irony is not lost on the platform: by using AI-powered technology to fight against AI-generated spam, they’re essentially fighting fire with fire.

According to Reddit’s own statistics, their updated tools have proven effective in reducing spam rates. They claim to block 23 million spam views per day and catch around 25,000 new spam posts and comments daily.

The company notes that these LLM-powered tools are able to detect subtle patterns of fake behavior and artificial hype that older systems often missed. This has led to a significant reduction in users’ exposure to spam – by about 20% from January to March compared with the prior three months.

Reddit’s approach is not unique, as other social platforms like YouTube, Meta, and Instagram have also implemented automated tools to reduce spam. However, Reddit’s updated system appears to be catching spam at a higher rate than its predecessors.

The use of AI-generated content has become increasingly common on these platforms, with some allowing users to post such content provided they disclose it as artificial. TikTok takes this a step further by giving users the option to toggle how much AI-generated content they want to see.

While detecting and flagging AI-generated content can be beneficial in reducing spam and hate speech, experts emphasize that human moderation is still crucial for effective results.