A new research paper is adding fresh heat to the growing debate over AI “distillation”, a technique used to transfer capabilities from one AI model to another. The issue has attracted particular attention as Chinese open-weight models such as Moonshot AI’s Kimi K3 continue to gain ground. Researchers say they have found a way to uncover hidden reasoning traces generated by advanced AI models before they produce their final answers. They also suggest that similar techniques could potentially be used to obtain reasoning information from powerful proprietary models and use it to train other systems through distillation.

The study was carried out by researchers from the University of Tübingen, the Max Planck Institute, MATS Research and cybersecurity firm Snyk. In their experiments, the researchers observed notable similarities between reasoning generated by Kimi K3 and hidden reasoning traces associated with Anthropic’s Claude Opus 4.8 and OpenAI’s GPT-5.6 Sol when responding to certain prompts. However, the researchers made clear that these similarities do not prove that Moonshot AI or another Chinese developer actually distilled reasoning data from those US models.

The paper specifically states that the results cannot establish a causal link. Researchers also found that the US-based open-weight model Inkling from Thinking Machines and a DeepSeek model did not show the same type of reasoning similarities with Claude Opus. The controversy comes as artificial intelligence development has become an increasingly important area of competition between the United States and China. Distillation itself is not new and has long been used in machine learning to transfer capabilities from larger models to smaller and more efficient systems. But the technique has become politically and commercially sensitive as companies accuse rivals of using their models to accelerate development.

OpenAI previously alleged that DeepSeek had used outputs from its models while developing its R1 reasoning system. Anthropic also accused Chinese technology giant Alibaba of similar activity. At the same time, some Silicon Valley leaders have defended distillation as a legitimate part of the wider AI ecosystem. Meta CEO Mark Zuckerberg recently argued that restricting the practice could place US developers at a disadvantage.

The researchers' method for examining hidden reasoning relies on an unusual property of smaller models. While proprietary systems generally conceal their chain-of-thought reasoning, an encrypted version of this information is transmitted to a user's computer for computational purposes. The researchers found that feeding these traces into a smaller model from the same family could cause it to reveal more of the underlying reasoning because smaller models typically undergo less alignment training.

To investigate whether open-weight models might show evidence of similar training, the researchers tested 90 questions across several systems. They then supplied the models with the opening portions of reasoning traces obtained from a proprietary model. According to the study, some models produced answers that showed similarities to the original reasoning, with Kimi K3 displaying particularly noticeable overlaps.

The researchers cautioned that these results should not be treated as proof that any particular company copied confidential reasoning data. The research also raises a separate cybersecurity concern. The researchers warned that techniques capable of exposing hidden model reasoning could potentially be abused to retrieve sensitive information supplied to AI systems, including passwords and API keys. They said tests involving frontier models from OpenAI, Anthropic and Google demonstrated that sensitive information could be recovered through API access. The companies were reportedly notified about the vulnerability and subsequently introduced measures to reduce the risk.

Anthropic spokesperson Michael Aciman told Wired that the company welcomed independent research into its models and had begun implementing short-term protections against the replay behaviour described in the study. He also denied that the researchers had accessed Anthropic's internal infrastructure or obtained encryption keys and personal data from its systems. The research therefore raises important questions about both AI security and competition. While the similarities between certain models are notable, the study does not establish that Chinese companies deliberately extracted and distilled proprietary reasoning from US AI systems. Further research will be needed to determine whether the observed overlaps result from distillation, common training data, model behaviour or other factors. As the AI race between the US and China intensifies, the ability to protect proprietary reasoning while allowing legitimate research and innovation is likely to become an increasingly important challenge for the industry.
Disclaimer: This image is taken from The Indian Express.

Google’s Gemini artificial intelligence assistant has crossed the 1 billion monthly active user mark, giving the company a major boost in the increasingly competitive global AI race. Google CEO Sundar Pichai announced the milestone on August 11, describing Gemini as the fastest-growing product in the company’s history. The achievement highlights the rapid shift in consumer behaviour as AI assistants become part of everyday digital life. Gemini’s user base has expanded sharply over the past year, helped by Google’s decision to integrate its AI technology across a wide range of products and services.

Google had reported around 400 million monthly active users for the Gemini app at its 2025 I/O developer conference. By May 2026, the figure had risen to more than 900 million, before approaching the billion-user mark in the following months. The latest milestone means Gemini has added hundreds of millions of users in a relatively short period.

The 1 billion figure specifically relates to monthly users of the Gemini app. Google also offers Gemini-powered features through products such as Search and Android, giving the company an even wider reach for its artificial intelligence technology. Gemini’s growth comes as Google continues to expand the capabilities of its AI assistant. The platform can handle text, images and voice interactions and is increasingly being developed to perform more complex, multi-step tasks. Google has been working to move Gemini beyond a traditional chatbot and make it a more capable digital assistant that can help users with research, writing, planning, learning and other activities.

The company’s existing ecosystem has also played an important role in Gemini’s expansion. Because Google operates some of the world’s most widely used internet services and the Android mobile operating system, it has numerous opportunities to put Gemini in front of consumers. The milestone puts Gemini firmly alongside other leading consumer AI platforms. OpenAI’s ChatGPT has also reached the billion-user level, intensifying the competition between the two companies. The battle is increasingly shifting from attracting users to keeping them engaged and convincing them to rely on AI assistants for more sophisticated tasks.

For Google, the challenge now is to turn Gemini’s enormous audience into sustained usage and, potentially, a significant source of revenue. The company has introduced paid Gemini plans with additional capabilities while continuing to offer free access to its AI assistant. The rapid growth also comes with greater responsibility. As more people use AI to obtain information, create content and make decisions, accuracy and reliability are becoming increasingly important. A mistake made by an AI system can have a much wider impact when the platform is used by a billion people each month.

Gemini’s latest milestone nevertheless represents a significant achievement for Google. The company has managed to turn its AI assistant into a mainstream consumer product in a remarkably short period. With artificial intelligence becoming increasingly embedded across Search, Android and other Google services, Gemini’s next phase could determine how central AI becomes to the company’s future. For now, crossing the 1 billion monthly user mark gives Google a powerful position in the global AI market and signals that the era of mass-market AI assistants is already well underway.
Disclaimer: This image is taken from Reuters.

A Mumbai-based defence start-up, Techno Defence Pvt Ltd, has developed an indigenous Joule-Thomson (JT) Cooler designed for missile seekers, marking a major advancement in India’s defence technology capabilities. The development could also create opportunities in the high-value international defence market, although exports of the technology remain subject to strict Missile Technology Control Regime (MTCR) rules, Business Line reported.

The technology was developed with support from the Defence Research and Development Organisation (DRDO), following an initiative that began several years ago. The project traces its origins to a meeting around 2017–18 between IIT-Bombay alumnus Dr Pravin Salinkar and a director of a DRDO laboratory. Salinkar, who was then 75, took up the challenge of developing a specialised cooling system that had traditionally been available from only a handful of Western countries.

He later joined forces with fellow IIT-Mumbai alumnus Sudarshan Saraf to establish Techno Defence Pvt Ltd. The company brought together a compact group of engineers and technicians, with women making up a significant part of the technical workforce. The project received financial assistance through DRDO’s Technology Development Fund (TDF), allowing the team to work on the design and manufacture of the miniature JT Cooler.

The device performs a crucial function inside missiles equipped with infrared seekers. It rapidly cools infrared detectors and focal plane arrays to extremely low temperatures, reaching roughly –196°C. Such cooling is necessary because infrared sensors operating at normal temperatures generate thermal noise, commonly referred to as dark current.

Excessive background noise can interfere with weak infrared signals emitted or reflected by distant targets. By cooling the sensor, the JT Cooler significantly reduces this noise, allowing the seeker to identify and track targets more accurately. The technology is considered strategically important and falls under MTCR controls. Until now, advanced versions of such systems have largely been associated with countries including the United States, France and Israel. While missile systems based on satellite navigation may not require this type of cooling, infrared-guided missiles depend heavily on it for effective seeker performance.

Techno Defence developed its own manufacturing approach rather than simply replicating existing foreign designs. The company worked with small-scale and cottage industries to develop specialised production methods and equipment needed for the system. Manufacturing the cooler involves highly precise processes, including microscopic assembly work, the use of high-purity materials and production in controlled environments. Each stage requires strict quality checks because even small manufacturing variations can affect the reliability of equipment intended for defence applications.

The project was completed during 2023–24 and subsequently handed over to DRDO for testing. The trials were successful, validating the performance of the indigenous system and strengthening India’s position in a highly specialised area of missile technology. The development has also reportedly generated interest among overseas buyers. However, any export of the technology would require the necessary government approvals because of MTCR-related restrictions.

The achievement is significant for India’s broader push towards defence self-reliance. Developing the JT Cooler domestically could reduce dependence on imported components while giving Indian industry an opportunity to participate in specialised global defence supply chains. Techno Defence operates as a joint venture with Technocraft Industries (India) Limited. Its work has also involved DRDO’s Solid State Physics Laboratory and Research Centre Imarat, bringing together private-sector manufacturing expertise and government research capabilities.

The project illustrates how a relatively small engineering team, supported by indigenous suppliers and government-backed funding, was able to develop a sophisticated component for missile systems. It also highlights the potential of India’s private defence sector to address complex technological requirements and eventually compete in specialised international markets. With initiatives such as the Technology Development Fund encouraging collaboration between DRDO and private companies, developments like the JT Cooler could play an increasingly important role in expanding India’s domestic defence manufacturing ecosystem and technological independence.
Disclaimer: This image is taken from Indian Defence News.

The global tablet market weakened in the second quarter of 2026, with shipments falling 10 per cent year-on-year to around 36 million units, according to research firm Omdia. The decline came as manufacturers faced continued component shortages while consumers became more cautious about spending because of higher device prices.

The drop was particularly notable because the April-June quarter typically benefits from seasonal demand linked to back-to-school purchases. However, tablet shipments failed to grow sequentially, suggesting that both supply constraints and weaker purchasing sentiment are weighing on the market. Omdia expects these challenges to persist through the rest of 2026, with affordable tablets likely to face the greatest pressure as manufacturers struggle to secure components.

Despite the slowdown, the research firm does not believe consumer interest in tablets has disappeared. Tablets continue to be widely used in households, schools and some business environments. Larger-screen devices remain useful for entertainment and everyday computing, while detachable tablets are increasingly being considered as alternatives to conventional laptops for certain tasks.

Omdia believes the current decline is largely the result of market conditions rather than a fundamental change in consumer behaviour. Limited component availability has reduced supplies of lower-cost devices, while rising prices have encouraged some buyers to delay purchases or hold on to existing tablets for longer. The supply situation could also influence the types of tablets manufacturers choose to produce. With components in short supply, vendors are expected to give greater priority to premium and flagship models, which generally provide better margins. This could make entry-level tablets less widely available and potentially push average selling prices higher.

Apple continued to lead the global tablet market during the quarter despite seeing its shipments fall by around 8 per cent year-on-year. The company shipped approximately 13.5 million iPads worldwide, giving it a 38 per cent share of the market. The standard iPad accounted for most of Apple's shipments, while the iPad Air and iPad mini contributed smaller volumes. Samsung remained in second place, shipping close to 6 million tablets during the quarter. Its shipments declined 13 per cent compared with the same period a year earlier. Omdia expects Samsung to adjust its product mix as component shortages continue, with greater emphasis likely to be placed on higher-margin models.

Lenovo was the standout performer among the major tablet manufacturers. Its shipments increased 27 per cent year-on-year, making it the only leading vendor to record growth during the quarter. Omdia said the increase was supported by actual consumer demand, while additional shipments to distribution channels ahead of expected price increases and major promotional events also helped boost the company's numbers. Xiaomi ranked fourth among the leading tablet vendors after shipping around 2.8 million units, representing a 7 per cent annual decline. The company continued to benefit from strong demand in China and steady growth in parts of the Asia-Pacific region. Huawei completed the top five, with shipments of approximately 2.7 million units, down 16 per cent from the previous year. Its tablet business remains largely concentrated in China, with limited availability in overseas markets.

The weakness in the wider personal computing market was also visible in the Chromebook segment. Omdia reported declining shipments as budget and education-focused devices faced increasing cost pressures. Memory has become a particularly significant part of the bill of materials for inexpensive Chromebooks, making the segment more vulnerable to supply and pricing challenges. Lenovo remained the largest Chromebook supplier, although its shipments slipped 3 per cent year-on-year to about 1.76 million units. The decline was largely linked to the completion of the first phase of Japan's GIGA 2.0 School Program, where Lenovo had been a major supplier.

Acer followed with 1.18 million units, down 4 per cent from a year earlier. Demand in North America and some market-share gains from HP helped support its performance. HP ranked third, with shipments falling 13 per cent to approximately 1.15 million units amid relatively weaker institutional demand in the education sector. Asus was the strongest performer in the Chromebook market. Its shipments jumped 66 per cent year-on-year to around 749,000 units, lifting its market share from 8 per cent to 16 per cent. Omdia attributed the increase to Asus's involvement in Japan's GIGA 2.0 programme, the US K–12 education replacement cycle and retailers building stock ahead of the back-to-school season.

Dell, meanwhile, recorded a sharp 47 per cent decline in Chromebook shipments, which fell to around 465,000 units. Omdia said weaker education demand has been an important factor behind the Chromebook slowdown. Some vendors are also placing less emphasis on the category because of its comparatively low profitability. The year-on-year comparison has been made more difficult by the unusually strong education-related shipments recorded during the same period in 2025.

Some previously planned education deployments have also been delayed in 2026, further reducing Chromebook volumes. Omdia expects the global tablet market to remain under pressure for the rest of the year. Component shortages, higher prices and delayed purchases are likely to continue affecting shipments, particularly in the budget segment. Manufacturers may increasingly concentrate their limited supplies on premium devices while exploring services, artificial intelligence features and new use cases to create additional sources of growth. The latest figures suggest that the tablet market is not necessarily losing its relevance. Instead, the industry is navigating a difficult period in which supply limitations and rising costs are influencing both what manufacturers produce and when consumers choose to buy.
Disclaimer: This image is taken from Magnific.

Meta’s Ray-Ban smart glasses have rapidly emerged as one of the world’s most popular new tech products, with reports suggesting that more than seven million pairs were sold in 2025. Supporters praise the glasses for making photography and accessibility more convenient, but the technology has also sparked privacy concerns. Critics have dubbed them “pervert glasses,” while some UK pubs and restaurants, including Wetherspoons, have reportedly banned customers from using the devices on their premises.

Disclaimer: This podcast is taken from The Guardian.

On the July 24 episode of Open For Business, hosts Andrea Heng and Poh Kok Ing are joined by Santosh Rao, Head of Research and Partner at Manhattan Venture Partners, to break down today's market trends.

Disclaimer: This podcast is taken from CNA.

As AI continues to evolve, cyber risks are becoming a major business challenge rather than just a technical problem. The Five Eyes alliance warns that advanced AI models could transform the cyber threat landscape faster than anticipated. With AI being used for both attacks and defense, the question remains: who is ahead in this new automated cyber battle? Andrea Heng and Hairianto Diman explore this with Jayant Dave, Chief Information Security Officer at Check Point Software Technologies.

Disclaimer: This podcast is taken from CNA.

A prolonged and heated courtroom dispute between tech billionaires Elon Musk and Sam Altman has ended in a win for OpenAI’s CEO. Musk says he plans to challenge the decision. The case has raised wider questions about Big Tech influence and the worldwide competition in artificial intelligence. Lucy Hough discusses the outcome with Guardian US tech and power reporter Nick Robins-Early in a YouTube interview.
Disclaimer: This image is taken from The Guardian.