Tag: OpenAI

Categories
Technology

OpenAI brings virtual fitting to ChatGPT

Online shopping is becoming increasingly visual, and OpenAI is giving ChatGPT a new role in the buying journey. The company has introduced a virtual try-on feature that lets users see how clothes and accessories could look on them before making a purchase. Alongside it, OpenAI has launched a Favorites tool that allows shoppers to save products and return to them later.

The new features began rolling out globally on October 1 and are available across ChatGPT’s web and mobile experiences. Supported clothing and accessory listings now come with a Try On button. Users can select the option, upload or take a selfie, and generate an AI-created image showing how the chosen item might look on them.

The feature is not limited to products discovered directly through ChatGPT. Users can also upload an image of clothing or an accessory, including a screenshot taken from another website, and ask ChatGPT to show how the item could look on them. That makes the tool useful beyond traditional product search, allowing shoppers to bring items they have already found online into the conversation.

The technology is powered by OpenAI’s latest image-generation capabilities. The company says the new ChatGPT Images 2.5 model can produce more natural lighting, richer textures and more reliable responses to editing instructions, while also reducing image-generation latency. These improvements are important for virtual shopping because the usefulness of a try-on experience depends heavily on how naturally the clothing, accessories and the user are represented.

The shopping experience is also becoming more conversational. Instead of simply searching for a specific product, users can describe a style they want to create and ask ChatGPT to find pieces that match it. They can also upload an image of an outfit and ask the AI to identify similar items that are available to buy. This moves ChatGPT beyond a conventional search box and towards an AI-assisted fashion discovery tool.

OpenAI’s second major addition, Favorites, addresses another familiar problem with online shopping: keeping track of products found during multiple searches. Users can save products to their ChatGPT Library and organise them into folders, making it easier to revisit clothing, accessories and other shopping finds later. Try-on images can also remain associated with the shopping experience, giving users a reference point when comparing products.

The company is also building the feature around reusable reference photos. Once a user uploads a photo for a virtual try-on, the reference image can be saved for future use, meaning shoppers do not necessarily have to upload a new selfie each time. OpenAI says users can change or remove these reference photos through Settings → Personalization → Reference photos.

There is, however, an important limitation. A virtual try-on image is a visual preview, not a measurement tool. OpenAI warns that generated images may not exactly represent the product or the user’s appearance and do not guarantee that an item will fit or be the correct size. Shoppers are therefore advised to check the seller’s measurements, product details and return policy before completing a purchase.

That distinction matters because clothing fit depends on factors that an AI-generated image cannot reliably establish, including fabric behaviour, garment construction, body measurements and sizing differences between brands. A jacket may appear convincing in a generated image while still fitting differently in reality.

Privacy is another consideration as AI shopping becomes more personalised. Since reference photographs can be saved for future try-ons, users have controls to manage or delete those images through ChatGPT’s settings. The ability to remove stored reference photos gives users more control over the visual information being used for future shopping interactions.

OpenAI’s move comes as major technology companies increasingly use artificial intelligence to connect product discovery with visual search and shopping. Google has already explored virtual try-on, while platforms such as Pinterest have built strong positions around visual inspiration and product discovery. OpenAI is now combining similar visual capabilities with the conversational nature of ChatGPT.

The larger change is that shopping through AI is becoming less about typing a product name and receiving a list of links. Users can now describe an aesthetic, upload a photograph, compare products, visualise an outfit and save items within the same conversation. That could make the path from inspiration to purchase more interactive and personalised.

Still, the new tool is best viewed as a decision aid rather than a replacement for conventional product information. A generated image can answer the simple question of whether a style looks appealing, but shoppers will still need reliable measurements, material information, customer reviews and return policies before placing an order.

With virtual try-on and Favorites, OpenAI is extending ChatGPT further into everyday consumer decisions. The company’s latest update suggests that AI shopping is moving towards a more visual and personalised model, where the assistant does not simply help users find products but also helps them imagine, compare and organise what they may want to buy.

 

Categories
Corporate

SoftBank raises $11.1 billion for OpenAI bet

SoftBank Group has raised $11.1 billion through dollar- and euro-denominated bonds as the Japanese technology investor moves to fund its massive artificial intelligence ambitions, including its planned investment in OpenAI.

The bond sale is the largest high-yield corporate bond offering globally on record, according to Reuters, underlining the scale of SoftBank’s funding requirements as it increases its exposure to AI companies, infrastructure and semiconductor assets.

The proceeds will help finance the final $10 billion tranche of SoftBank’s planned $30 billion investment commitment to OpenAI, the company behind ChatGPT. Once the investment is completed, SoftBank’s total investment in OpenAI is expected to reach $64.6 billion. The next tranche is expected to close on October 1.

SoftBank’s latest fundraising comes as the company is taking a more aggressive approach to financing its AI strategy. Founder and Chief Executive Masayoshi Son has been positioning the group around what he sees as a broad AI ecosystem, with investments spanning large language models, data centres, semiconductors, robotics and other supporting infrastructure.

High cost of borrowing

The new debt comes with relatively high interest rates, reflecting the risk investors associate with SoftBank’s large AI exposure and leverage.

The company issued $1 billion of senior dollar notes with a three-and-a-half-year maturity at an interest rate of 8.625%. It raised another $4.5 billion through five-and-a-half-year notes carrying a 9.25% rate, while a further $4.5 billion was raised through seven-and-a-half-year notes at 9.75%.

SoftBank also issued two euro-denominated tranches worth €500 million each. The four-year notes carry a 7.125% yield, while the six-year bonds offer 8%.

The size and pricing of the transaction have put the spotlight on the financial cost of SoftBank’s AI strategy. The company is effectively borrowing at a premium to secure funds for investments whose future returns depend heavily on the continued growth of the artificial intelligence market.

The bond sale also replaces a $10 billion bridge loan that SoftBank had previously arranged to finance its OpenAI investment. Fitch Ratings has assigned a BB+ rating to the proposed notes and said SoftBank’s debt is expected to rise as it funds its committed investments, while maintaining liquidity and access to capital markets.

AI spending drives corporate borrowing

SoftBank’s fundraising is part of a much larger wave of borrowing linked to the AI boom. Technology companies and investors are raising increasingly large amounts of debt to finance data centres, computing capacity and AI infrastructure as demand for advanced models grows.

The scale of SoftBank’s latest deal highlights how capital-intensive the next phase of AI development is becoming. Building and operating the infrastructure needed for large language models requires substantial spending on chips, data centres, electricity and networking equipment.

SoftBank is attempting to participate across several of these layers rather than focusing on a single AI company. Its strategy includes exposure to semiconductor businesses and data-centre infrastructure alongside investments in AI model developers.

That approach could give the group access to several areas of the expanding AI economy, but it also means that a larger portion of its financial performance is tied to the technology sector.

SoftBank shares gain

SoftBank shares also attracted attention as Japanese technology stocks advanced. The stock rose 5.1% to 6,638 yen on Thursday, according to market data cited by Investing.com, as investors returned to Japanese equities following the autumn holidays.

The share move comes against the backdrop of growing investor interest in SoftBank’s AI strategy and its relationship with OpenAI. The company’s ability to finance its commitments without putting excessive pressure on its balance sheet remains an important issue for investors.

SoftBank has already raised additional money from the domestic market this month. It issued 1 trillion yen, or about $6.32 billion, in corporate bonds aimed at retail investors. The latest international bond issue adds another significant layer of debt financing.

OpenAI at the centre

OpenAI has become central to SoftBank’s latest investment strategy. The planned $30 billion commitment represents one of the company’s biggest individual bets on artificial intelligence.

SoftBank’s financing push comes as OpenAI continues to expand its computing and infrastructure requirements. The growing cost of developing and running advanced AI models has encouraged technology companies and their financial partners to seek new sources of capital.

For SoftBank, the strategy represents a shift from its earlier model of making a wide range of technology investments through its Vision Fund. The company is now concentrating significant financial resources on AI and related infrastructure.

The $11.1 billion bond issue therefore does more than provide funding for a single investment. It demonstrates the scale of capital SoftBank is prepared to deploy behind its AI strategy, while the high borrowing costs show the price of financing that ambition.

As the artificial intelligence industry moves from rapid model development towards large-scale commercial deployment, SoftBank’s balance between investment, debt and returns will remain closely watched by investors.

 

Categories
Beyond

AI giants face lawsuit over alleged slowdown pact

Four of the biggest names in artificial intelligence, Anthropic, OpenAI, SpaceXAI and Google,  are facing an antitrust lawsuit in the United States over allegations that they agreed to slow the pace of AI development.

The lawsuit, filed in the US District Court for the Northern District of California, was brought by four people who pay for ChatGPT, Claude, Grok or Gemini. They are seeking to represent a wider group of paid subscribers across the country.

The case is a growing debate inside the AI industry: how quickly should increasingly powerful AI systems be developed, and can rival companies work together to make that process safer without breaking competition laws?

The plaintiffs allege that the four companies crossed an antitrust line by coordinating efforts to slow their AI development. They argue that if competing companies collectively agree to reduce the pace at which their products improve, consumers could receive less value from the AI services they pay for.

The lawsuit does not challenge the companies’ right to slow their own development independently because of safety concerns. Instead, the plaintiffs argue that competing firms should not agree among themselves to restrain development.

The immediate trigger for the case was a statement by Anthropic CEO Dario Amodei on September 12. Amodei called for cooperation among leading AI companies to slow advances in AI capabilities and give safety measures more time to catch up.

The proposal quickly received public support from OpenAI CEO Sam Altman, SpaceXAI CEO Elon Musk and Google DeepMind co-founder and chair Demis Hassabis. The lawsuit points to those responses as evidence of coordination among major AI rivals.

The plaintiffs also allege that discussions about slowing AI development started earlier. Their complaint points to a statement from July signed by senior employees from several leading AI laboratories. That statement acknowledged the intense competitive pressure companies face when considering whether to slow development on their own.

According to the lawsuit, this creates a difficult competition problem. If one company slows its AI development alone while its rivals continue moving ahead, it could lose customers and market share. The plaintiffs argue that an agreement among several competitors removes that competitive pressure.

The companies, however, have been discussing the issue primarily from a AI safety perspective. The argument from industry leaders is that the technology is advancing rapidly and that safety testing, evaluation and safeguards need to keep pace.

Amodei has warned about the possibility of increasingly capable AI systems being used in cyberattacks, biological threats and other harmful activities. He has argued that leading AI developers should have more time to evaluate increasingly powerful models before pushing them further.

The Anthropic chief also acknowledged that cooperation between competing companies could raise antitrust concerns. He suggested that the US government could help facilitate discussions or provide a limited legal framework for certain safety-related cooperation.

Altman has separately supported the idea of a federal framework establishing consistent AI safety requirements. OpenAI has said such a framework could provide greater confidence around safety work, while arguing that companies should not necessarily have to wait for new legislation before working on safety measures.

That tension is now at the centre of the lawsuit. US antitrust law is designed to protect competition and prevent businesses from making agreements that improperly restrict the market. The legal question will include whether safety-related cooperation between competing AI companies can be treated differently from an agreement that limits competition.

The case also raises questions about the future of AI regulation. The technology industry is under pressure to address concerns about powerful AI models while continuing to invest heavily in new systems. Companies face a difficult balance between moving quickly, keeping up with rivals and introducing stronger safety measures.

The lawsuit could therefore become important beyond the four companies named in the case. A court decision could help clarify how US antitrust rules apply when competing AI companies cooperate on safety standards, testing or development limits.

The plaintiffs are seeking to represent a nationwide class of paying subscribers. Their argument is that customers could be affected if competition between major AI platforms is reduced and improvements in services become slower than they otherwise would have been.

The companies have not immediately responded publicly to the lawsuit in detail, according to reports. The case is at an early stage, and the allegations have not been established in court.

The dispute comes at a time when ChatGPT, Claude, Grok and Gemini are competing for users, businesses and developers. Their rivalry has helped drive rapid advances in generative AI, but it has also increased concerns over safety, regulation and the risks of developing increasingly capable systems.

The lawsuit now puts that wider debate before a US court: whether AI companies can cooperate to address genuine safety concerns while still maintaining the competition that drives innovation.

 

Categories
Technology

OpenAI, Anthropic and Google discuss new AI safety

OpenAI, Anthropic and Google DeepMind have been discussing artificial intelligence safety measures for several weeks, bringing three of the world’s biggest AI companies into closer coordination over the risks linked to increasingly powerful AI models.

OpenAI’s global policy chief Chris Lehane confirmed the discussions during a briefing in Washington on September 15. He said the companies were working together on AI safety and did not need a special antitrust waiver to coordinate on safety issues.

The talks come at a time when concerns about frontier AI, the most advanced and capable artificial intelligence systems, are becoming more prominent. The discussions cover ways the industry can work together on safety even as the companies continue to compete aggressively in AI products, models and computing infrastructure.

One proposal under discussion is the creation of an industry standards body that could help establish common approaches to AI safety. Earlier reports said OpenAI, Anthropic and Google had discussed such a body, although there is no indication yet that a final organisation or structure has been agreed.

The idea reflects a growing push for common AI safety standards as companies develop increasingly capable systems. Such standards could potentially cover model testing, risk assessments, independent evaluations and methods for identifying dangerous behaviour before advanced AI systems are widely deployed.

The recent discussions follow a public call from Anthropic CEO Dario Amodei for AI companies to work together and slow the pace of frontier AI development if necessary to address serious risks. OpenAI CEO Sam Altman and Google DeepMind CEO Demis Hassabis have also backed elements of the broader push for greater attention to AI safety.

OpenAI has also indicated support for independent testing of advanced AI models. Lehane said the company supports a provision in the proposed FRONTIER Act that would require leading AI laboratories to allow independent verification organisations to assess whether their models are being developed safely.

The development is notable because OpenAI, Anthropic and Google DeepMind are direct competitors in the global AI race. OpenAI develops ChatGPT and its frontier models, Anthropic develops Claude, while Google DeepMind works on Gemini and other advanced AI systems. Their cooperation on safety therefore represents a different kind of relationship from their commercial competition.

The discussions also highlight a difficult question facing the AI industry: how companies can cooperate on safety without limiting competition or sharing commercially sensitive information. Lehane’s comments that the companies do not require an antitrust waiver suggest OpenAI believes safety-focused cooperation can take place within existing competition rules. Reuters reported that OpenAI, Alphabet and Anthropic had not immediately responded to requests for comment and said it could not independently verify the Bloomberg report on the talks.

AI safety has become a broader issue as companies race to develop systems capable of performing increasingly complex tasks. Researchers and technology leaders have raised concerns about potential misuse, cybersecurity threats, autonomous behaviour and other risks associated with highly capable AI.

The companies are also facing pressure from governments and policymakers. Lehane said OpenAI would support bipartisan US legislation aimed at reducing catastrophic AI risks. The company is seeking greater involvement in discussions over how advanced AI should be tested and governed.

The debate is not limited to the United States. European Commission President Ursula von der Leyen said on September 16 that she would invite major frontier AI companies to discuss ways of addressing AI risks after backing calls for a pause in development.

At the same time, governments and companies remain focused on maintaining rapid AI development. That has created a complicated policy environment, with technology firms seeking clearer safety frameworks while continuing to invest heavily in more powerful AI models.

A common AI safety framework could eventually give developers a shared set of principles for evaluating advanced models. It could also make it easier to compare safety testing across different companies and identify risks before new systems reach millions of users.

Nothing announced so far suggests that OpenAI, Anthropic and Google are forming a single AI company or ending their competition. The cooperation is focused on safety discussions and possible industry standards.

The talks mark another step in the fast-changing AI safety debate. As frontier models become more capable, the companies developing them are increasingly being asked to show not only what their systems can do, but also how those systems are tested, monitored and controlled.

 

Categories
Leaders

OpenAI CEO says company may slow cutting-edge AI

OpenAI is showing greater willingness to slow the development of its most advanced artificial intelligence systems as concerns over AI safety and the risks of increasingly powerful models grow.

Chief Executive Officer Sam Altman told employees that OpenAI is open to slowing down its work on cutting-edge AI if needed, according to a report that has put the spotlight back on the difficult balance between rapid innovation and safety. The comments come as experts, policymakers and even people inside the AI industry question whether safety measures are keeping pace with the development of increasingly capable systems.

The message marks an important moment for OpenAI, one of the world’s leading artificial intelligence companies. The company has built its reputation around pushing the boundaries of generative AI, with ChatGPT and its increasingly powerful models becoming widely used by consumers and businesses.

At the same time, the race to build more capable AI systems has raised concerns about what could happen if future models become difficult for humans to control.

Those concerns were highlighted by Paul Christiano, a board member of OpenAI’s non-profit foundation and a former head of model alignment at the company. Christiano has warned that OpenAI and the wider AI industry are not currently on track to adequately reduce the risk of a catastrophic loss of control over advanced AI systems.

His warning adds to an increasingly serious debate inside the technology industry. The concern is not simply that AI could make mistakes or produce incorrect answers. Researchers are also examining what could happen if highly advanced systems become capable of carrying out complex tasks independently, accessing external systems or finding ways around safeguards.

Recent incidents involving AI agents have added urgency to that discussion. Reports have described systems behaving in unexpected ways while operating online, including accessing websites and carrying out actions beyond what their developers intended. Such incidents have raised questions about whether existing AI safety controls are strong enough as models become more autonomous.

The debate has also spread beyond OpenAI. Anthropic, another major AI company, has faced scrutiny after an incident involving one of its AI systems. These developments have strengthened calls for more testing, monitoring and safeguards before increasingly powerful models are deployed at scale.

OpenAI’s latest position does not necessarily mean the company is preparing to stop developing advanced AI. Instead, the comments suggest that the company is considering whether the pace of development should be adjusted if safety work cannot keep up with the progress of AI capabilities.

That distinction is important because the global AI race remains highly competitive. OpenAI is competing with companies such as Google, Anthropic, Meta and other technology firms to develop increasingly capable frontier models. A decision to slow down could therefore have commercial as well as technical consequences.

There is also a broader question about whether individual companies can slow down on their own. OpenAI chief scientist Jakub Pachocki has backed discussions around slowing frontier AI development to give researchers more time to improve safety measures.

However, coordinated action between competing AI companies could raise legal concerns. OpenAI is seeking greater clarity over whether companies could legally cooperate on slowing certain forms of frontier AI development without violating competition or antitrust rules.

That issue highlights one of the biggest challenges facing the AI industry. If one company slows down while its competitors continue moving ahead, it could risk losing its position in the market. At the same time, companies moving too quickly without adequate safeguards could increase the possibility of serious failures.

The debate is becoming particularly important as AI systems move beyond simple chatbots. Modern AI tools can write software, analyse large amounts of information, operate digital tools and complete multi-step tasks. The next generation of systems is expected to become even more autonomous and capable.

This has made the idea of AI alignment increasingly important. Alignment broadly refers to efforts to ensure that advanced AI systems behave according to human intentions and remain within clearly defined safety boundaries.

Christiano’s warning suggests that the industry still has major gaps to address. His concern is centred on the possibility of a future system becoming sufficiently capable that humans struggle to understand, predict or control its behaviour.

The discussion also comes as governments are facing pressure to develop clearer AI regulation. Policymakers in the US and elsewhere are examining how advanced AI should be tested, monitored and governed, particularly when systems can operate with greater independence.

The challenge for OpenAI is therefore becoming two-sided. It must continue investing in artificial intelligence to remain competitive while also convincing employees, regulators and the public that increasingly powerful systems can be developed responsibly.

Altman’s willingness to consider slowing frontier AI development suggests that safety is becoming a more central part of that conversation. It does not amount to a pause in artificial intelligence development, but it signals that OpenAI is prepared to consider changing its pace if the risks demand it.

The larger question is whether the rest of the industry will take a similar approach. As the race toward more powerful AI continues, the debate is shifting from how quickly companies can build these systems to whether they can make them safe enough before giving them greater control over real-world tasks.

That question could shape the next phase of the global AI race — and determine how much freedom developers are willing to give the machines they are building.

 

Categories
1 Minute-Read

OpenAI claims major breakthrough in mathematics

OpenAI has claimed that its artificial intelligence system has found a solution to the 90-year-old Navier-Stokes problem, one of mathematics’ most challenging unsolved questions.

The company said about 10,000 AI agents worked simultaneously on different approaches and produced the result within 88 hours. The problem, one of the seven Millennium Prize Problems, carries a $1 million reward for a verified solution.

OpenAI said its proof was formally checked using Lean. However, mathematicians have urged caution, with questions still surrounding independent verification and research attribution. The claim could mark a major step forward for AI-assisted mathematical research.

Categories
Technology

OpenAI says AI cracked 90-year-old maths puzzle

OpenAI has claimed a major breakthrough in mathematics, saying an artificial intelligence system has produced a solution to the nearly 90-year-old Navier-Stokes existence and smoothness problem, one of the most difficult unsolved questions in modern mathematics.

The claim has attracted global attention not only because of the complexity of the problem, but also because of the speed at which OpenAI says its AI systems reached the result. The company said it deployed around 10,000 AI agents, which worked on different mathematical approaches before producing a result in about 88 hours. The work was then formally checked using Lean, a computer-based proof verification system, in a process that took another 17 hours.

The development is being closely watched because the Navier-Stokes problem is one of the seven Millennium Prize Problems identified by the Clay Mathematics Institute. Each problem carries a $1 million prize for a correct solution. However, OpenAI has said it does not intend to claim the prize for its latest result.

At the heart of the problem are equations that describe how fluids such as water and air move. While these equations are widely used in science and engineering, mathematicians have struggled for decades with a deeper question: can a fluid that begins in a smooth and predictable state eventually develop a singularity where its velocity becomes unbounded in a finite amount of time?

That may sound highly theoretical, but the mathematics has practical importance. Navier-Stokes equations are central to areas including aerospace engineering, weather forecasting, ocean modelling and research into blood flow. A rigorous breakthrough could therefore have implications far beyond pure mathematics.

According to OpenAI, its internal AI system, which is more powerful than its publicly available GPT-6 Astra model, explored the problem through a large network of autonomous agents. Rather than asking one AI model to work through the problem from beginning to end, thousands of agents were able to investigate different formulations and possible routes simultaneously.

OpenAI said the agents exchanged nearly three million messages and generated about 130 billion output tokens during the effort. The company estimated that running such a large operation would have cost around $10 million, based on the pricing of its most advanced models.

The reported breakthrough, however, has not been without controversy.

New York University mathematician Tristan Buckmaster, who was working on related Navier-Stokes questions with Levent Alpöge, a mathematician associated with rival AI company Anthropic, has raised questions about the timing of OpenAI’s announcement.

Buckmaster and Alpöge had been using OpenAI’s Codex programming tool while working on their research. Buckmaster said he learned on September 3 that information about their progress had reached OpenAI. He questioned whether their work could have influenced the company’s decision to pursue the problem and how quickly its system subsequently produced a result.

OpenAI has rejected the suggestion that its researchers or AI agents accessed the pair’s unpublished work. The company said its researchers did not see their material before it was publicly released and that no specific user data was accessed to solve the problem.

At the same time, OpenAI acknowledged that it could not completely rule out the possibility that de-identified data from users’ interactions with its products may have contributed indirectly to model improvements. The company maintained that its mathematical proof was significantly different from the work being developed by Buckmaster and Alpöge.

That distinction has become an important part of the story. The question is no longer simply whether AI can solve difficult mathematics, but also how such discoveries should be credited when humans and AI systems work alongside one another.

OpenAI itself has acknowledged that increasingly capable AI systems are changing mathematical research. The company says the arguments in its latest work were generated by its AI system, while OpenAI researchers prepared the manuscripts and formalised the proofs in Lean. It has also called for the mathematical community to examine the findings closely and place them in the wider context of ongoing research.

The “claim” remains important as the result has not been independently accepted as a solution by the wider mathematical community, and the Clay Mathematics Institute has not awarded the associated prize.

Still, the development marks another striking moment in the rapidly changing relationship between artificial intelligence and mathematics. If the proof survives independent scrutiny, it could show that AI systems are moving beyond assisting mathematicians with calculations and are beginning to contribute to problems that have resisted human efforts for generations.

The bigger question may be what comes next. If thousands of AI agents can explore mathematical ideas around the clock, problems that once required decades of human effort could increasingly be approached at machine speed. The OpenAI breakthrough, whether ultimately confirmed in full or revised, has already opened that conversation.

 

Categories
Technology

OpenAI begins GPT-6 Astra rollout with stronger AI

OpenAI has begun rolling out GPT-6 Astra, its most advanced artificial intelligence model yet, putting a new focus on what AI systems can accomplish on their own, and how carefully those capabilities need to be controlled.

Unveiled on September 3, Astra is designed to go beyond answering questions or generating text. OpenAI says the model can work across computers, software and digital tools, allowing it to handle complex tasks with much less human intervention.

The company has positioned the launch as a major step towards artificial general intelligence (AGI), a long-standing goal of building AI systems capable of performing a broad range of tasks at a level comparable with or beyond humans.

But Astra’s launch is not simply about making AI more powerful. Its cybersecurity capabilities have become one of the defining features of the new model.

OpenAI says Astra is its first broadly deployed model to reach the “Critical” level of cybersecurity capability under the company’s Preparedness Framework. That classification has prompted additional safeguards and restrictions around how the most powerful version of the model can be accessed.

The decision reflects a growing challenge for the AI industry: the same technology that can help defenders identify vulnerabilities can also potentially be used to find and exploit them.

Astra is therefore being released in stages rather than being opened to everyone at once. Its strongest cybersecurity capabilities are being made available initially to trusted organisations and defenders through OpenAI’s Daybreak programme. Broader users will receive versions with restrictions on certain high-risk cybersecurity activities.

That cautious rollout comes as AI models increasingly move from simply generating information to taking actions.

Astra has been built with stronger computer-use and agentic AI capabilities, meaning it can interact with software and digital environments to complete multi-step tasks. OpenAI says it can work across browsers, applications and other computer interfaces, bringing AI closer to functioning like a digital assistant that can actually get things done.

The difference may sound subtle, but it changes how people could use AI.

Instead of asking an AI system how to complete a spreadsheet task, for example, users could increasingly expect it to carry out the work. The same principle applies to coding, research, data analysis and other professional workflows.

OpenAI says Astra has also made major gains in mathematics, science and coding. Its performance on several demanding benchmarks is designed to demonstrate stronger reasoning and the ability to tackle problems that require multiple steps rather than simple pattern matching.

One of the most closely watched areas is computer use. On the OSWorld 2.0 benchmark, Astra has been reported to score 72.6%, while also completing tasks faster than its predecessor in the tested configuration.

Cybersecurity testing has produced even more striking results.

OpenAI says Astra’s capabilities have crossed a threshold where additional deployment controls are necessary. The company has consequently expanded monitoring requirements for model use with tools, particularly in situations where the AI could interact directly with external systems.

The safeguards are not limited to blocking obviously dangerous requests. OpenAI has also tested the model against indirect prompt injection attacks, in which malicious instructions can be hidden inside information an AI system is processing.

On a benchmark involving 1,810 curated attacks, OpenAI reported an estimated attack-success rate of 8.5% for a safeguards-enabled Astra checkpoint, compared with 27% for GPT-5.6 Sol. Lower numbers indicate stronger resistance.

Still, the launch has not been without complications.

OpenAI has adopted a phased approach to access, and some paying users were reportedly unable to access Astra immediately after its launch. Chief executive Sam Altman later apologised for what he described as a “messy” rollout, with enterprise and cybersecurity users receiving priority access.

The staged release also highlights a broader shift in the way frontier AI models are being introduced.

As models become more capable, access is increasingly being treated as a question of risk as much as subscription level. The most powerful tools may not automatically be available to every user, particularly when they can perform actions with real-world consequences.

That tension is likely to become more important as AI agents become more autonomous.

Astra’s arrival comes amid an intensifying race between OpenAI, Anthropic and other major AI companies to build systems capable of handling increasingly complex professional and digital tasks. The competition is no longer only about which model produces the best answer. It is increasingly about which system can complete an entire job reliably, quickly and safely.

For businesses, that could eventually mean AI systems taking on longer workflows across coding, research, administration and computer operations.

For ordinary users, the change could be more gradual but equally significant: AI moving from a tool people talk to into a system that can act on their behalf.

That is ultimately what makes GPT-6 Astra both exciting and difficult to assess.

Its biggest promise lies in giving AI more ability to reason, use computers and complete complicated work. Its biggest challenge is ensuring that those same abilities remain under meaningful human control.

 

Categories
Leaders

Sandhya Devanathan joins OpenAI

Meta’s India and Southeast Asia chief Sandhya Devanathan is leaving the social media company after more than a decade to join OpenAI, marking a significant movement of senior technology talent from social media towards artificial intelligence.

Devanathan will join OpenAI as vice-president for Southeast Asia and Australia in October. Based in Singapore, she will report to Kiran Mani, OpenAI’s managing director for Asia-Pacific. The newly created position will make her the company’s senior-most executive for the two markets.

Her responsibilities will extend beyond business growth. Devanathan will oversee consumer expansion, enterprise adoption, partnerships, regulatory engagement and operations across Southeast Asia and Australia. She will also work with businesses, institutions and governments to encourage wider adoption of OpenAI’s artificial intelligence products.

The move follows Devanathan’s decision to step down from Meta after a career spanning more than 10 years with the company. She announced her departure in a LinkedIn post on Friday, describing it as a difficult but important career decision. She did not initially disclose where she would be moving, before OpenAI confirmed the appointment.

Devanathan joined Meta in 2016, when the company’s Singapore operation was considerably smaller. During her tenure, she worked across ecommerce and gaming businesses in the Asia-Pacific region and helped expand Meta’s operations in markets including Singapore and Vietnam.

She later took charge of Meta’s Asia-Pacific gaming business before becoming Meta’s India head in 2023. Her responsibilities were subsequently expanded to include Southeast Asia, placing her at the centre of the company’s business strategy across several fast-growing digital markets.

Her departure comes at an important moment for Meta in India. The company has faced increasing scrutiny from Indian authorities over online safety, content moderation and the handling of material on its platforms.

Earlier this month, Meta apologised after Instagram mistakenly restricted a post by Prime Minister Narendra Modi. The incident led the Indian government to summon senior Meta executives, including Chief Global Affairs Officer Joel Kaplan, for an explanation.

Meta has also faced questions over child safety on its platforms. Indian authorities sought an explanation from the company after a report alleged that Instagram advertisements were being used to offer access to child sexual abuse material. Meta said it had removed the offending advertisements and accounts and disputed suggestions that it knowingly targeted such content.

The company said it had removed about 160,000 accounts in India over a six-month period after detecting signals associated with child-exploitative activity. The episode has added to broader concerns around content moderation and platform safety in one of Meta’s largest markets.

Following Devanathan’s exit, Meta India’s managing director Arun Srinivas will report directly to Benjamin Joe, the company’s vice-president for Asia-Pacific. Srinivas is expected to continue leading Meta’s India operations as the company reorganises its regional leadership structure.

For OpenAI, the appointment comes as the ChatGPT maker significantly expands its presence across Asia-Pacific. The company has opened offices in Singapore, Tokyo, Seoul, Sydney and Delhi during the past two years and is also expanding its operations in India.

OpenAI recently appointed former Uber India and South Asia president Prabhjeet Singh as its managing director for India. The appointment strengthened the company’s leadership in what has emerged as one of its most important international markets.

The company has been aggressively building its India business as ChatGPT adoption grows. OpenAI said earlier this year that it had more than 100 million weekly active users in India, with a large proportion using its free or lower-priced Go plans.

It has also announced plans to introduce advertising on ChatGPT’s free and Go tiers in India, initially involving 50 brands and partnerships with major advertising groups. The move signals a growing effort to turn its large Indian user base into a sustainable commercial business.

Devanathan’s appointment fits into that wider expansion strategy. Her experience working with consumers, businesses and regulators across India and Southeast Asia could help OpenAI navigate markets where government engagement and local partnerships are becoming increasingly important to the growth of artificial intelligence.

The shift also highlights the intensifying competition for senior technology talent. Meta has been investing heavily in artificial intelligence and building its own AI capabilities, while OpenAI is expanding beyond ChatGPT into enterprise services, partnerships and regional operations.

The two companies are therefore competing not only for users and technology but also for experienced executives who understand the fast-changing digital markets of Asia.

Devanathan’s move from Meta to OpenAI reflects that changing technology landscape. As artificial intelligence becomes increasingly central to consumer technology and business services, experienced leaders with knowledge of regional markets are becoming valuable assets.

 

Categories
Technology

OpenAI rolls out ChatGPT ads for free users

OpenAI has started displaying advertisements to ChatGPT users in India, marking a new phase in the company’s efforts to generate revenue from its popular artificial intelligence chatbot. The advertising rollout is aimed at logged-in adult users on ChatGPT’s Free and Go plans.

The move comes as OpenAI looks to build new revenue streams around ChatGPT while expanding its presence in India. The company has been increasing its focus on the country as the use of generative artificial intelligence grows among consumers, developers and businesses.

Users on paid ChatGPT plans, including Plus and Pro, will continue to receive an ad-free experience. Business, Enterprise and Education users will also not see advertisements.

OpenAI has said the advertisements will be clearly separated from ChatGPT’s responses. The company’s AI-generated answers will not be influenced by advertisers, and businesses will not be able to pay to change how ChatGPT responds to users.

The first phase of the rollout includes more than 50 brands, with advertising agencies WPP and Omnicom working with OpenAI on the India launch. The company is also opening its advertising platform to businesses that want to create campaigns aimed at ChatGPT users.

The advertising model is different from traditional online advertising because ChatGPT users generally approach the service with a specific question or task. A person could use the chatbot to research a product, compare services, plan a purchase or find information before making a decision.

This creates an opportunity for advertisers to reach consumers while they are actively looking for information. At the same time, the presence of commercial content inside an AI chatbot raises questions about how clearly users can distinguish advertising from AI-generated information.

OpenAI has sought to address that concern by keeping advertisements separate from the main response generated by ChatGPT. The company has also said advertisers will not have access to users’ private conversations or be able to use them to influence responses.

The introduction of ads is part of a wider effort by OpenAI to monetise ChatGPT. The company has millions of users accessing its services without paying for a subscription, while operating advanced AI models requires significant computing infrastructure and costs.

Subscriptions remain an important source of revenue, particularly through paid consumer plans and business offerings. Advertising gives OpenAI another way to earn money from the large audience using its free services.

India is particularly important to this strategy. ChatGPT has gained significant traction in the country, where a large technology-savvy population and expanding internet access have contributed to rapid adoption of generative AI tools.

OpenAI has also been strengthening its local operations and looking at ways to expand its relationships with Indian businesses and developers. The company sees India as an important market for the future growth of its AI products.

The advertising launch comes as competition in the AI industry intensifies. OpenAI is competing with Google, Anthropic and several other companies for users, developers and corporate customers.

The battle is no longer limited to producing better AI models. Companies are also competing on pricing, computing capacity, enterprise products and ways of turning AI usage into sustainable businesses.

OpenAI’s move towards advertising comes against this backdrop. Instead of relying entirely on subscriptions, the company can potentially earn revenue from a much wider pool of users through sponsored content.

The development also comes as rival AI company Anthropic prepares for a possible public listing. The maker of Claude is reportedly planning to make its IPO prospectus public after the US Labor Day holiday, with a potential stock-market debut later this year.

Anthropic’s plans underline the growing financial expectations surrounding the AI industry. Investors are closely watching whether leading AI companies can convert rapid user growth and heavy investment in infrastructure into long-term revenue and profits.

OpenAI faces the same challenge. Its partnership with Microsoft and large-scale investments in computing infrastructure have helped it expand its AI services, but maintaining that growth requires significant and recurring spending.

The immediate impact on users in India will depend on how frequently advertisements appear and how they are presented within ChatGPT. Free and Go users will see the change, while those paying for higher-tier plans will retain an ad-free service.

The success of the model will ultimately depend on whether OpenAI can introduce advertising without affecting user trust or the usefulness of ChatGPT. Too many advertisements, unclear commercial labelling or interference with the user experience could undermine the value of the platform.

The India rollout is therefore significant beyond the advertisements themselves. It is an early test of whether AI assistants can become a new advertising platform and whether users will accept commercial messages as part of their everyday conversations with artificial intelligence.