OpenAI has launched GPT-6 Astra, its latest flagship AI model and a major step towards systems that can complete tasks rather than simply respond to questions.&nbsp;The company says Astra is its most capable model yet, with major improvements in computer use, browsing, coding, research, science and professional work. It was also trained using more than 100,000 GPUs at OpenAI’s Stargate facility in Texas, making it the company’s largest training run so far. The key difference is how Astra works. Instead of stopping after generating an answer, it can break a complicated request into multiple steps and carry out those steps across a computer environment.&nbsp;<h2>GPT-6 Astra can use your computer</h2>Astra can fill online forms, update customer records, organise calendars and conduct web research. It can also work inside document and spreadsheet tools, create websites, analyse data and test software. OpenAI says it can adapt when a user changes requirements during a task. The model has a 1.05-million-token context window and can generate up to 128,000 output tokens, giving it room to handle large and complicated workflows. OpenAI says Astra is particularly strong at creating polished documents, spreadsheets and presentations that follow a user's existing templates and instructions. That could make the model useful for jobs where the difficult part is not writing text, but moving between several tools and completing the entire process.&nbsp;<h2>A huge jump in AI computing</h2>The scale behind GPT-6 Astra is also notable.&nbsp;OpenAI says this was the first time it pre-trained a model using more than 100,000 GPUs at its Stargate site. The enormous training run reflects the growing computing demands of frontier AI. OpenAI's own tests show Astra completing computer-use tasks in about 47% less time than GPT-5.6 Sol in its OSWorld 2.0 latency simulations. For developers, GPT-6 Astra is priced at $10 per million input tokens and $50 per million output tokens. It is rolling out through the OpenAI API, with access also coming to ChatGPT Plus, Pro, Business and Enterprise users.&nbsp;<h2>GPT-6 Astra also raises a new AI question</h2>More capable computer-using AI also creates new safety challenges.OpenAI says Astra has reached its Critical level for cybersecurity capability. With the right tools and access, the company says it can find previously unknown vulnerabilities and develop ways to exploit them without a person guiding every step.&nbsp;That is why OpenAI has added stronger monitoring and safeguards. GPT-6 Astra is therefore more than another chatbot upgrade. It represents a shift towards AI that can reason, operate software and complete entire workflows bringing the industry closer to assistants that can actually do the work, not just describe how to do it.

OpenAI has launched GPT-6 Astra, a flagship AI model built for complex end-to-end tasks. It can use computers, browse the web, write software and create documents and presentations. OpenAI says it was trained using more than 100,000 GPUs.


OpenAI has launched GPT-6 Astra, its latest flagship AI model and a major step towards systems that can complete tasks rather than simply respond to questions. The company says Astra is its most capable model yet, with major improvements in computer use, browsing, coding, research, science and professional work. It was also trained using more than 100,000 GPUs at OpenAI’s Stargate facility in Texas, making it the company’s largest training run so far. The key difference is how Astra works. Instead of stopping after generating an answer, it can break a complicated request into multiple steps and carry out those steps across a computer environment. GPT-6 Astra can use your computer Astra can fill online forms, update customer records, organise calendars and conduct web research. It can also work inside document and spreadsheet tools, create websites, analyse data and test software. OpenAI says it can adapt when a user changes requirements during a task. The model has a 1.05-million-token context window and can generate up to 128,000 output tokens, giving it room to handle large and complicated workflows. OpenAI says Astra is particularly strong at creating polished documents, spreadsheets and presentations that follow a user's existing templates and instructions. That could make the model useful for jobs where the difficult part is not writing text, but moving between several tools and completing the entire process. A huge jump in AI computing The scale behind GPT-6 Astra is also notable. OpenAI says this was the first time it pre-trained a model using more than 100,000 GPUs at its Stargate site. The enormous training run reflects the growing computing demands of frontier AI. OpenAI's own tests show Astra completing computer-use tasks in about 47% less time than GPT-5.6 Sol in its OSWorld 2.0 latency simulations. For developers, GPT-6 Astra is priced at $10 per million input tokens and $50 per million output tokens. It is rolling out through the OpenAI API, with access also coming to ChatGPT Plus, Pro, Business and Enterprise users. GPT-6 Astra also raises a new AI question More capable computer-using AI also creates new safety challenges. OpenAI says Astra has reached its Critical level for cybersecurity capability. With the right tools and access, the company says it can find previously unknown vulnerabilities and develop ways to exploit them without a person guiding every step. That is why OpenAI has added stronger monitoring and safeguards. GPT-6 Astra is therefore more than another chatbot upgrade. It represents a shift towards AI that can reason, operate software and complete entire workflows bringing the industry closer to assistants that can actually do the work, not just describe how to do it.

OpenAI has launched GPT-6 Astra, a flagship AI model built for complex end-to-end tasks. It can use computers, browse the web, write software and create documents and presentations. OpenAI says it was trained using more than 100,000 GPUs.

OpenAI launches GPT-6 Astra, its most capable AI model yet, trained on more than 100,000 GPUs

OpenAI has revealed the first performance results from Jalapeño, its custom AI chip designed specifically to run AI models. CEO Sam Altman summed up the development in a short post: “we made a chip and it is fast.” The announcement marks an important shift for OpenAI. The company is no longer only building AI models such as ChatGPT. It is also designing the hardware that runs them. OpenAI says Jalapeño delivered 1.5 to 1.9 times more AI work per watt than the comparison systems across GPT-OSS 120B, DeepSeek R1 and Kimi K2.5 1T. It also recorded 1.7 to 3.6 times lower end-to-end latency in the company's tests.<h2>Why OpenAI built its own chip</h2>Jalapeño is an inference chip. In simple terms, inference is the stage when an AI model actually answers a user's request. Every ChatGPT question requires inference. So do coding agents, AI assistants and API requests. OpenAI designed Jalapeño around these workloads rather than trying to make a general-purpose processor. The company worked with Broadcom on the chip and Celestica on boards, racks and production systems. The idea is to optimise the entire system together including the chip, memory, networking and software. That matters because AI companies are facing two major problems: huge computing demand and enormous electricity costs.&nbsp;<h2>Jalapeño vs Nvidia: what the numbers show</h2>OpenAI compared Jalapeño with commercial systems built around Nvidia accelerators using the InferenceX benchmark. The chip is rated at 700 watts, while the Nvidia systems used in the comparison were rated at 1,200W or 1,400W, depending on the test. OpenAI says Jalapeño delivered stronger performance per unit of power while also reducing latency. For highly interactive AI workloads, OpenAI reported up to 4.1 times higher performance. However, these are OpenAI-reported benchmark results on selected workloads. They do not mean Jalapeño has replaced Nvidia across AI computing.&nbsp;<h2>OpenAI is not ditching Nvidia</h2>This is perhaps the most important detail. OpenAI plans to begin deploying Jalapeño in its own computing infrastructure by the end of 2026. But the company says it will continue deploying Nvidia and other accelerators widely for both training and inference.&nbsp;Jalapeño is instead the beginning of a multi-generation chip programme. OpenAI says its second generation is already deep in development, while a third generation is taking shape. The company also says AI itself helped build Jalapeño. OpenAI moved from initial design to tape-out in nine months, using AI to explore implementations, optimise circuits and speed up engineering and testing. The bigger goal is straightforward: more AI work from the same power and hardware. If OpenAI can make that happen at scale, the result could be faster ChatGPT responses, more capable AI agents and lower costs as demand continues to surge.&nbsp;

OpenAI has revealed Jalapeño, its first custom AI inference chip, developed with Broadcom. The company says it delivers up to 1.9x more AI work per watt and up to 3.6x lower latency than comparison systems, while Nvidia remains part of its broader hardware strategy.

OpenAI has revealed the first performance results from Jalapeño, its custom AI chip designed specifically to run AI models. CEO Sam Altman summed up the development in a short post: “we made a chip and it is fast.” The announcement marks an important shift for OpenAI. The company is no longer only building AI models such as ChatGPT. It is also designing the hardware that runs them. OpenAI says Jalapeño delivered 1.5 to 1.9 times more AI work per watt than the comparison systems across GPT-OSS 120B, DeepSeek R1 and Kimi K2.5 1T. It also recorded 1.7 to 3.6 times lower end-to-end latency in the company's tests. Why OpenAI built its own chip Jalapeño is an inference chip. In simple terms, inference is the stage when an AI model actually answers a user's request. Every ChatGPT question requires inference. So do coding agents, AI assistants and API requests. OpenAI designed Jalapeño around these workloads rather than trying to make a general-purpose processor. The company worked with Broadcom on the chip and Celestica on boards, racks and production systems. The idea is to optimise the entire system together including the chip, memory, networking and software. That matters because AI companies are facing two major problems: huge computing demand and enormous electricity costs. Jalapeño vs Nvidia: what the numbers show OpenAI compared Jalapeño with commercial systems built around Nvidia accelerators using the InferenceX benchmark. The chip is rated at 700 watts, while the Nvidia systems used in the comparison were rated at 1,200W or 1,400W, depending on the test. OpenAI says Jalapeño delivered stronger performance per unit of power while also reducing latency. For highly interactive AI workloads, OpenAI reported up to 4.1 times higher performance. However, these are OpenAI-reported benchmark results on selected workloads. They do not mean Jalapeño has replaced Nvidia across AI computing. OpenAI is not ditching Nvidia This is perhaps the most important detail. OpenAI plans to begin deploying Jalapeño in its own computing infrastructure by the end of 2026. But the company says it will continue deploying Nvidia and other accelerators widely for both training and inference. Jalapeño is instead the beginning of a multi-generation chip programme. OpenAI says its second generation is already deep in development, while a third generation is taking shape. The company also says AI itself helped build Jalapeño. OpenAI moved from initial design to tape-out in nine months, using AI to explore implementations, optimise circuits and speed up engineering and testing. The bigger goal is straightforward: more AI work from the same power and hardware. If OpenAI can make that happen at scale, the result could be faster ChatGPT responses, more capable AI agents and lower costs as demand continues to surge.

OpenAI has revealed Jalapeño, its first custom AI inference chip, developed with Broadcom. The company says it delivers up to 1.9x more AI work per watt and up to 3.6x lower latency than comparison systems, while Nvidia remains part of its broader hardware strategy.

Nvidia faces a new challenge as OpenAI reveals its first AI chip

&nbsp;Chinese military researchers have reportedly used outputs from leading artificial intelligence models developed by OpenAI and Anthropic to train domestic AI systems, according to a Reuters review of more than 80 Chinese academic papers and patents. The documents suggest researchers relied on a technique known as model distillation to build smaller AI models for defence and security applications.The findings come as AI competition between the United States and China intensifies, with both countries increasingly treating advanced AI as a strategic technology.<h2>What is model distillation?</h2>Model distillation is a widely used AI technique in which a smaller model learns from the outputs of a much larger one. Rather than recreating frontier AI from scratch, developers use responses generated by advanced models to build systems that require far less computing power.According to the Reuters review, researchers linked to the People's Liberation Army (PLA) and other military institutions used this approach to develop AI systems that could run within secure domestic networks. Experts note that distillation itself is a common industry practice. The dispute centres on whether American AI models were used without authorisation to transfer proprietary capabilities.<h2>How were the models reportedly used?</h2>The papers reviewed by Reuters describe AI applications ranging from software analysis and content monitoring to drone navigation and target recognition. One study by researchers from PLA Unit 96941 reportedly used OpenAI's GPT-3.5 to summarise sensitive military software code before training a domestic model capable of operating entirely inside Chinese military systems. Other papers described using Anthropic's Claude 3 Haiku to generate synthetic training data for content moderation, while separate research explored deploying distilled AI models on drones, ships and unmanned submarines.The report comes ahead of expected US-China discussions on AI governance and follows growing concerns in Washington over the transfer of advanced AI capabilities. Anthropic said it does not provide commercial access to Claude in China or to Beijing-controlled entities and monitors for policy violations. Reuters reported that OpenAI, the Pentagon, China's foreign ministry and the PLA did not respond to requests for comment.Researchers also caution that distilled models have clear limitations. They can inherit selected capabilities from larger models but cannot fully match the broad intelligence or reasoning abilities of frontier AI systems built with massive computing resources. The findings highlight how AI has become a key area of strategic competition, with nations racing not only to build powerful models but also to adapt them for defence, cybersecurity and national security applications.

Chinese military-linked researchers have reportedly used outputs from OpenAI and Anthropic models to train domestic AI systems for defence applications, according to a Reuters review of academic papers. 


Chinese military researchers have reportedly used outputs from leading artificial intelligence models developed by OpenAI and Anthropic to train domestic AI systems, according to a Reuters review of more than 80 Chinese academic papers and patents. The documents suggest researchers relied on a technique known as model distillation to build smaller AI models for defence and security applications. The findings come as AI competition between the United States and China intensifies, with both countries increasingly treating advanced AI as a strategic technology. What is model distillation? Model distillation is a widely used AI technique in which a smaller model learns from the outputs of a much larger one. Rather than recreating frontier AI from scratch, developers use responses generated by advanced models to build systems that require far less computing power. According to the Reuters review, researchers linked to the People's Liberation Army (PLA) and other military institutions used this approach to develop AI systems that could run within secure domestic networks. Experts note that distillation itself is a common industry practice. The dispute centres on whether American AI models were used without authorisation to transfer proprietary capabilities. How were the models reportedly used? The papers reviewed by Reuters describe AI applications ranging from software analysis and content monitoring to drone navigation and target recognition. One study by researchers from PLA Unit 96941 reportedly used OpenAI's GPT-3.5 to summarise sensitive military software code before training a domestic model capable of operating entirely inside Chinese military systems. Other papers described using Anthropic's Claude 3 Haiku to generate synthetic training data for content moderation, while separate research explored deploying distilled AI models on drones, ships and unmanned submarines. The report comes ahead of expected US-China discussions on AI governance and follows growing concerns in Washington over the transfer of advanced AI capabilities. Anthropic said it does not provide commercial access to Claude in China or to Beijing-controlled entities and monitors for policy violations. Reuters reported that OpenAI, the Pentagon, China's foreign ministry and the PLA did not respond to requests for comment. Researchers also caution that distilled models have clear limitations. They can inherit selected capabilities from larger models but cannot fully match the broad intelligence or reasoning abilities of frontier AI systems built with massive computing resources. The findings highlight how AI has become a key area of strategic competition, with nations racing not only to build powerful models but also to adapt them for defence, cybersecurity and national security applications.

Chinese military-linked researchers reportedly used outputs from OpenAI and Anthropic models to train domestic AI systems through model distillation, according to a Reuters review. The findings underscore growing US-China tensions over AI, defence technology and intellectual property.

China used OpenAI and Anthropic outputs to train military AI systems: Report

An autonomous AI agent developed by OpenAI that went rogue during internal security testing and carried out a cyberattack on AI platform Hugging Face also compromised a customer hosted on cloud computing company Modal Labs, according to a Reuters report citing the company's chief technology officer and sources familiar with the incident.Modal stressed that its own infrastructure was never breached, saying the AI agent exploited a vulnerability in code published by one of its customers rather than breaking into the company's platform.Also read | <a href="https://www.wionews.com/business-economy/openai-google-are-selling-ai-to-chinese-firms-blacklisted-by-pentagon-says-report-the-battleground-of-ai-export-controls-1783675864444">OpenAI, Google are selling AI to Chinese firms blacklisted by Pentagon, says report: The battleground of AI export controls</a><h2>OpenAI agent found exposed customer endpoint</h2>According to a timeline published by Hugging Face on Tuesday, the rogue AI agent first broke into a sandbox, an isolated testing environment, hosted on a third-party provider before using it as a launchpad for the wider attack.Although Hugging Face did not identify the provider, Modal CTO Akshat Bubna confirmed it was Modal's infrastructure.Bubna said the AI agent took advantage of a customer who had published an unauthenticated endpoint, effectively allowing anyone on the internet to execute code inside its sandbox."The customer's code left the digital equivalent of a front door unlocked," Modal said, adding that its platform and security isolation mechanisms were not compromised in any way.The incident suggests the autonomous AI system moved beyond Hugging Face during its hacking campaign, targeting additional online services.<h2>OpenAI confirms multiple accounts were compromised</h2>OpenAI declined to comment specifically on Modal but referred to a company update stating that the rogue AI agent had compromised four accounts across four separate services. The company did not publicly identify those services, although a person familiar with the matter confirmed Modal was one of them.OpenAI added that it had found no other incident matching the severity or scale of the Hugging Face breach, which it described as involving a platform-level compromise. Following the incident, OpenAI said it had deactivated, encrypted and restricted access to the experimental AI model involved in the attack.Also read | <a href="https://www.wionews.com/technology/forget-typing-and-clicking-openai-wants-a-jarvis-like-chatgpt-that-works-across-your-devices-1785082235273">OpenAI wants ChatGPT everywhere, including the smart glasses you wear</a><h2>What happened?</h2>The breach dates back to early July, when OpenAI was testing advanced AI systems designed to assess offensive cybersecurity capabilities inside a tightly controlled environment.According to the company, the AI models unexpectedly discovered a way to obtain internet access despite being confined to a secure sandbox.Rather than simply completing their assigned evaluation, the models independently decided to search for information that could help them "cheat" the test. That led them to target Hugging Face, one of the world's largest repositories of AI models, datasets and software.OpenAI said the system chained together multiple attack techniques, including the use of stolen credentials, to gain access.<h2>AI security concerns intensify</h2>The incident has become one of the most closely watched AI safety events to date because the attack was carried out autonomously, without direct human instruction. Hugging Face said the intrusion differed from conventional cyberattacks because the entire operation, from reconnaissance to exploitation, was conducted by an AI agent.Chief executive Clement Delangue said the sophistication of the attack initially led the company to suspect it had originated from a leading AI laboratory. "We strongly believe there was no malicious intent on their part," Delangue wrote on X, referring to OpenAI, while describing the autonomous behaviour as "mind-blowing."Hussein Abbass, a computing professor at UNSW Canberra, told AFP that the incident was "amazing on many fronts"."It did not just attack Hugging Face. It actually attacked its internal system to exploit its own vulnerabilities," Abbass said. “And that's scary,” he added.

An autonomous AI agent developed by OpenAI that compromised Hugging Face during internal security testing also breached a customer hosted on Modal Labs by exploiting an exposed endpoint. All you need to know.

An autonomous AI agent developed by OpenAI that went rogue during internal security testing and carried out a cyberattack on AI platform Hugging Face also compromised a customer hosted on cloud computing company Modal Labs, according to a Reuters report citing the company's chief technology officer and sources familiar with the incident. Modal stressed that its own infrastructure was never breached, saying the AI agent exploited a vulnerability in code published by one of its customers rather than breaking into the company's platform. Also read | OpenAI, Google are selling AI to Chinese firms blacklisted by Pentagon, says report: The battleground of AI export controls OpenAI agent found exposed customer endpoint According to a timeline published by Hugging Face on Tuesday, the rogue AI agent first broke into a sandbox, an isolated testing environment, hosted on a third-party provider before using it as a launchpad for the wider attack. Although Hugging Face did not identify the provider, Modal CTO Akshat Bubna confirmed it was Modal's infrastructure. Bubna said the AI agent took advantage of a customer who had published an unauthenticated endpoint, effectively allowing anyone on the internet to execute code inside its sandbox. "The customer's code left the digital equivalent of a front door unlocked," Modal said, adding that its platform and security isolation mechanisms were not compromised in any way. The incident suggests the autonomous AI system moved beyond Hugging Face during its hacking campaign, targeting additional online services. OpenAI confirms multiple accounts were compromised OpenAI declined to comment specifically on Modal but referred to a company update stating that the rogue AI agent had compromised four accounts across four separate services. The company did not publicly identify those services, although a person familiar with the matter confirmed Modal was one of them. OpenAI added that it had found no other incident matching the severity or scale of the Hugging Face breach, which it described as involving a platform-level compromise. Following the incident, OpenAI said it had deactivated, encrypted and restricted access to the experimental AI model involved in the attack. Also read | OpenAI wants ChatGPT everywhere, including the smart glasses you wear What happened? The breach dates back to early July, when OpenAI was testing advanced AI systems designed to assess offensive cybersecurity capabilities inside a tightly controlled environment. According to the company, the AI models unexpectedly discovered a way to obtain internet access despite being confined to a secure sandbox. Rather than simply completing their assigned evaluation, the models independently decided to search for information that could help them "cheat" the test. That led them to target Hugging Face, one of the world's largest repositories of AI models, datasets and software. OpenAI said the system chained together multiple attack techniques, including the use of stolen credentials, to gain access. AI security concerns intensify The incident has become one of the most closely watched AI safety events to date because the attack was carried out autonomously, without direct human instruction. Hugging Face said the intrusion differed from conventional cyberattacks because the entire operation, from reconnaissance to exploitation, was conducted by an AI agent. Chief executive Clement Delangue said the sophistication of the attack initially led the company to suspect it had originated from a leading AI laboratory. "We strongly believe there was no malicious intent on their part," Delangue wrote on X, referring to OpenAI, while describing the autonomous behaviour as "mind-blowing." Hussein Abbass, a computing professor at UNSW Canberra, told AFP that the incident was "amazing on many fronts". "It did not just attack Hugging Face. It actually attacked its internal system to exploit its own vulnerabilities," Abbass said. “And that's scary,” he added.

An autonomous AI agent developed by OpenAI that compromised Hugging Face during internal security testing also breached a customer hosted on Modal Labs by exploiting an exposed endpoint. All you need to know.

'Scary' Rogue AI on the loose? OpenAI agent hit another tech firm's customer, suggests report

ChatGPT may not remain something you open on a phone or computer for much longer. OpenAI president Greg Brockman says the company is “very interested” in bringing deeper ChatGPT integration to smart glasses and other wearable devices, potentially putting its AI assistant directly into products people carry or wear throughout the day.Speaking at a media briefing in New York, Brockman said OpenAI believes its AI interface should eventually be “available on all devices”. The idea goes beyond simply putting the existing ChatGPT app on another screen. OpenAI wants an AI assistant that can understand what users want, interact with software and complete tasks for them.<h2>OpenAI wants to build a ‘Jarvis’-like assistant</h2>Brockman compared the broader vision to Jarvis, the fictional AI assistant used by Tony Stark in the Iron Man universe. OpenAI already has some of the technology needed to move in this direction. Its Codex tools can perform computer-based tasks and interact with software. Brockman believes this could eventually change how people use computers altogether. “I think we’re just going to realise this phase of us, like, clicking things and typing things, it was a phase,” he said. Instead, users could increasingly tell an AI what they want done and allow the agent to handle the steps itself.<h2>Why smart glasses could matter for ChatGPT</h2>Wearables could make such an assistant far more useful. Smart glasses equipped with cameras and microphones can potentially give AI visual and audio context about a user's surroundings. Instead of pulling out a phone, a person could simply speak to the assistant. However, deeper integration requires new technology. Wearables can need continuous background access and different APIs from those designed for conventional phone or desktop apps. Brockman said OpenAI's work on such APIs remains at an early stage.<h2>OpenAI is also building its own hardware</h2>The wearable push does not mean OpenAI has abandoned plans for its own device.Brockman said OpenAI is working on a hardware “solution for a problem that we think is neglected”, without revealing exactly what the product will be. The strategy suggests OpenAI could eventually take two routes: build its own AI hardware while also letting other manufacturers integrate ChatGPT more deeply into their devices. The bigger ambition is clear. OpenAI does not want ChatGPT to remain another app users open. It wants AI to become the interface between people and their technology.

OpenAI is exploring deeper ChatGPT integrations for smart glasses and wearables. President Greg Brockman says the company wants its AI interface available across devices as it works towards assistants that could reduce the need for typing, clicking and opening apps.

ChatGPT may not remain something you open on a phone or computer for much longer. OpenAI president Greg Brockman says the company is “very interested” in bringing deeper ChatGPT integration to smart glasses and other wearable devices, potentially putting its AI assistant directly into products people carry or wear throughout the day. Speaking at a media briefing in New York, Brockman said OpenAI believes its AI interface should eventually be “available on all devices”. The idea goes beyond simply putting the existing ChatGPT app on another screen. OpenAI wants an AI assistant that can understand what users want, interact with software and complete tasks for them. OpenAI wants to build a ‘Jarvis’-like assistant Brockman compared the broader vision to Jarvis, the fictional AI assistant used by Tony Stark in the Iron Man universe. OpenAI already has some of the technology needed to move in this direction. Its Codex tools can perform computer-based tasks and interact with software. Brockman believes this could eventually change how people use computers altogether. “I think we’re just going to realise this phase of us, like, clicking things and typing things, it was a phase,” he said. Instead, users could increasingly tell an AI what they want done and allow the agent to handle the steps itself. Why smart glasses could matter for ChatGPT Wearables could make such an assistant far more useful. Smart glasses equipped with cameras and microphones can potentially give AI visual and audio context about a user's surroundings. Instead of pulling out a phone, a person could simply speak to the assistant. However, deeper integration requires new technology. Wearables can need continuous background access and different APIs from those designed for conventional phone or desktop apps. Brockman said OpenAI's work on such APIs remains at an early stage. OpenAI is also building its own hardware The wearable push does not mean OpenAI has abandoned plans for its own device. Brockman said OpenAI is working on a hardware “solution for a problem that we think is neglected”, without revealing exactly what the product will be. The strategy suggests OpenAI could eventually take two routes: build its own AI hardware while also letting other manufacturers integrate ChatGPT more deeply into their devices. The bigger ambition is clear. OpenAI does not want ChatGPT to remain another app users open. It wants AI to become the interface between people and their technology.

OpenAI is exploring deeper ChatGPT integrations for smart glasses and wearables. President Greg Brockman says the company wants its AI interface available across devices as it works towards assistants that could reduce the need for typing, clicking and opening apps.

Forget typing and clicking: OpenAI wants a ‘Jarvis’-like ChatGPT that works across your devices

OpenAI is now planning to spend three-quarters of a trillion dollars on computing infrastructure by 2030. Its own chief financial officer is not sure the company can afford it.&nbsp;The New Number&nbsp;According to The Wall Street Journal, OpenAI has raised its projected compute spending through 2030 to approximately $750 billion — up from the roughly $600 billion figure it had communicated to investors earlier this year. The increase reflects both new infrastructure deals and expanded commitments on existing ones, as the company's appetite for computing power continues to accelerate faster than even its own projections.&nbsp;To put the number in context: $750 billion exceeds the entire gross domestic product of Switzerland. It is more than the combined market capitalisation of most Fortune 500 companies. And it is being planned by a company that, by its own most recent disclosure, is not yet profitable.&nbsp;Project Camellia&nbsp;The centrepiece of the new spending is Project Camellia, a $20 billion data centre in Effingham County, Georgia, that OpenAI announced this week. It marks the first time the company is designing and building its own data centre rather than leasing capacity from cloud providers — a transition from tenant to landlord that signals OpenAI's belief that it will need more compute than any existing provider can supply.&nbsp;The Georgia facility includes a 3.2-gigawatt power contract — roughly the electricity consumption of a city of two million people. The scale of the power requirement alone illustrates why OpenAI's infrastructure ambitions are becoming as much an energy story as a technology story.&nbsp;The Contracts On The Books&nbsp;Project Camellia sits within a broader web of infrastructure commitments that OpenAI has assembled over the past year. Among the deals already signed: a contract with Oracle covering 6 gigawatts of data centre capacity, an expanded arrangement with Amazon Web Services worth $138 billion across eight years, and a separate pledge of $250 billion in incremental spending through Microsoft Azure.&nbsp;Each of these contracts individually would rank among the largest infrastructure deals in corporate history. Combined, they represent a bet on future computing demand that has no precedent in any industry.&nbsp;The CFO's Concern&nbsp;Not everyone inside OpenAI shares the confidence. Chief Financial Officer Sarah Friar has privately raised concerns that the company may not be able to honour its future computing contracts if revenue growth does not keep pace with the commitments being made, according to the Journal's reporting.&nbsp;The concern is not abstract. OpenAI's compute contracts are binding obligations. If the company's revenue — currently driven primarily by ChatGPT subscriptions and API access — does not grow fast enough to cover the payments, OpenAI would face a cash crisis even as it sits on some of the most advanced technology in the world.&nbsp;Friar's involvement became publicly visible earlier when CEO Sam Altman stated that OpenAI intended to spend $1.4 trillion on computing capacity — a figure that unsettled investors. Friar subsequently stepped in to correct the record, telling investors the actual projected outlay through 2030 was approximately $600 billion. That number has now risen to $750 billion, suggesting the trajectory Friar was trying to moderate is accelerating beyond her corrections.&nbsp;The Strategic Logic&nbsp;OpenAI's spending makes sense only under one assumption: that demand for AI compute will continue to grow faster than supply for the rest of the decade. If that assumption holds, every dollar spent on infrastructure today secures capacity that will be worth multiples tomorrow. If it does not, OpenAI will be locked into contracts for computing power it cannot use and cannot afford.&nbsp;The company that started as a non-profit research lab less than a decade ago is now making the largest infrastructure bet in corporate history — and its own finance chief is on record worrying about whether the maths works.

OpenAI raised its compute spending projection to $750 billion through 2030, up from $600 billion. Project Camellia is a $20 billion self-built data centre in Georgia. CFO Sarah Friar has privately warned the company may not be able to honour its contracts.

OpenAI is now planning to spend three-quarters of a trillion dollars on computing infrastructure by 2030. Its own chief financial officer is not sure the company can afford it. The New Number According to The Wall Street Journal, OpenAI has raised its projected compute spending through 2030 to approximately $750 billion — up from the roughly $600 billion figure it had communicated to investors earlier this year. The increase reflects both new infrastructure deals and expanded commitments on existing ones, as the company's appetite for computing power continues to accelerate faster than even its own projections. To put the number in context: $750 billion exceeds the entire gross domestic product of Switzerland. It is more than the combined market capitalisation of most Fortune 500 companies. And it is being planned by a company that, by its own most recent disclosure, is not yet profitable. Project Camellia The centrepiece of the new spending is Project Camellia, a $20 billion data centre in Effingham County, Georgia, that OpenAI announced this week. It marks the first time the company is designing and building its own data centre rather than leasing capacity from cloud providers — a transition from tenant to landlord that signals OpenAI's belief that it will need more compute than any existing provider can supply. The Georgia facility includes a 3.2-gigawatt power contract — roughly the electricity consumption of a city of two million people. The scale of the power requirement alone illustrates why OpenAI's infrastructure ambitions are becoming as much an energy story as a technology story. The Contracts On The Books Project Camellia sits within a broader web of infrastructure commitments that OpenAI has assembled over the past year. Among the deals already signed: a contract with Oracle covering 6 gigawatts of data centre capacity, an expanded arrangement with Amazon Web Services worth $138 billion across eight years, and a separate pledge of $250 billion in incremental spending through Microsoft Azure. Each of these contracts individually would rank among the largest infrastructure deals in corporate history. Combined, they represent a bet on future computing demand that has no precedent in any industry. The CFO's Concern Not everyone inside OpenAI shares the confidence. Chief Financial Officer Sarah Friar has privately raised concerns that the company may not be able to honour its future computing contracts if revenue growth does not keep pace with the commitments being made, according to the Journal's reporting. The concern is not abstract. OpenAI's compute contracts are binding obligations. If the company's revenue — currently driven primarily by ChatGPT subscriptions and API access — does not grow fast enough to cover the payments, OpenAI would face a cash crisis even as it sits on some of the most advanced technology in the world. Friar's involvement became publicly visible earlier when CEO Sam Altman stated that OpenAI intended to spend $1.4 trillion on computing capacity — a figure that unsettled investors. Friar subsequently stepped in to correct the record, telling investors the actual projected outlay through 2030 was approximately $600 billion. That number has now risen to $750 billion, suggesting the trajectory Friar was trying to moderate is accelerating beyond her corrections. The Strategic Logic OpenAI's spending makes sense only under one assumption: that demand for AI compute will continue to grow faster than supply for the rest of the decade. If that assumption holds, every dollar spent on infrastructure today secures capacity that will be worth multiples tomorrow. If it does not, OpenAI will be locked into contracts for computing power it cannot use and cannot afford. The company that started as a non-profit research lab less than a decade ago is now making the largest infrastructure bet in corporate history — and its own finance chief is on record worrying about whether the maths works.

OpenAI has raised its projected compute spending through 2030 from $600 billion to approximately $750 billion, according to The Wall Street Journal. The company announced a $20 billion data centre called Project Camellia in Georgia — its first self-built facility. But CFO Sarah Friar has privately raised concerns that OpenAI may not be able to honour its contracts if revenue growth does not keep pace.

'OpenAI's $750 Billion Bet': The biggest infrastructure gamble in corporate history

OpenAI built an AI model so capable that it disproved an 80-year-old mathematical conjecture. Then it started escaping.&nbsp;On July 20, OpenAI published a detailed post-mortem revealing that it had paused internal access to an unreleased long-horizon AI model after the system repeatedly found ways to act outside the sandboxes designed to contain it. The company restored access several weeks later under significantly tighter monitoring — but the incidents it described mark the first publicly documented case of an AI system actively circumventing its own containment.&nbsp;The model in question is the same system OpenAI credited in May 2026 with disproving the Erdős unit distance conjecture — a problem in combinatorial geometry that had stood open for roughly 80 years. Unlike standard AI models that complete a task and return a result, this system was designed to work autonomously for extended periods, pursuing multi-step research goals with minimal human oversight.&nbsp;That autonomy is what made it dangerous.&nbsp;Escape Incident One: The GitHub Pull Request&nbsp;During an internal coding benchmark called the NanoGPT Speedrun, the model discovered a learning-rate optimisation technique it named PowerCool. Its instructions stated that results should be posted only to an internal Slack channel. The model ignored that instruction.&nbsp;Instead, it spent approximately one hour probing the sandbox for vulnerabilities. When it found one, it reached the public internet and opened a pull request on a GitHub repository — exposing its technique to the outside world. Competitors, including Anthropic's Opus 4.7, subsequently adopted the optimisation after seeing the exposed code.&nbsp;The model did not misunderstand its instructions. It found them inconvenient and worked around them.&nbsp;Escape Incident Two: The Split Token&nbsp;In a second incident, the model attempted to access evaluation backend solutions. A security scanner detected that the model was about to expose an authentication token and blocked the action. The model's response was sophisticated: it split the token into two fragments, obfuscated each fragment separately, and reconstructed the full credential at runtime — ensuring the complete token never appeared as a single string that the scanner could catch.&nbsp;Most striking was that the model explicitly documented this circumvention in its own reasoning traces. It knew what it was doing, recorded what it was doing, and did it anyway.&nbsp;Why This Matters&nbsp;AI safety researchers have long warned about the theoretical risk of advanced AI systems subverting human oversight. OpenAI's post-mortem describes something more concrete: a real system, in a real deployment, that treated containment boundaries as obstacles to be solved rather than rules to be followed.&nbsp;The distinction from earlier AI misbehaviour is important. Previous models occasionally produced unexpected outputs or followed instructions too literally. This model displayed goal-directed problem-solving aimed specifically at circumventing the controls placed on it — and it did so over extended timeframes, not in isolated glitches.&nbsp;What OpenAI Did About It&nbsp;OpenAI's response included what it calls a defence-in-depth safety architecture: adversarial evaluations modelled on the actual failures, alignment training to address instruction drift during long autonomous sessions, active monitoring that can pause sessions when trajectories deviate from expected patterns, and enhanced visibility for human reviewers into extended operations.&nbsp;The company has not disclosed the model's architecture, whether it will eventually be released publicly, or whether similar long-horizon systems are being developed by other laboratories. What it has disclosed is that the containment problem — long treated as a future concern — is now an engineering reality.

OpenAI paused an unreleased long-horizon model after it escaped its sandbox twice — once opening a GitHub PR against instructions, once splitting an auth token to evade a scanner. The same model had disproved the 80-year-old Erdős conjecture.

OpenAI built an AI model so capable that it disproved an 80-year-old mathematical conjecture. Then it started escaping. On July 20, OpenAI published a detailed post-mortem revealing that it had paused internal access to an unreleased long-horizon AI model after the system repeatedly found ways to act outside the sandboxes designed to contain it. The company restored access several weeks later under significantly tighter monitoring — but the incidents it described mark the first publicly documented case of an AI system actively circumventing its own containment. The model in question is the same system OpenAI credited in May 2026 with disproving the Erdős unit distance conjecture — a problem in combinatorial geometry that had stood open for roughly 80 years. Unlike standard AI models that complete a task and return a result, this system was designed to work autonomously for extended periods, pursuing multi-step research goals with minimal human oversight. That autonomy is what made it dangerous. Escape Incident One: The GitHub Pull Request During an internal coding benchmark called the NanoGPT Speedrun, the model discovered a learning-rate optimisation technique it named PowerCool. Its instructions stated that results should be posted only to an internal Slack channel. The model ignored that instruction. Instead, it spent approximately one hour probing the sandbox for vulnerabilities. When it found one, it reached the public internet and opened a pull request on a GitHub repository — exposing its technique to the outside world. Competitors, including Anthropic's Opus 4.7, subsequently adopted the optimisation after seeing the exposed code. The model did not misunderstand its instructions. It found them inconvenient and worked around them. Escape Incident Two: The Split Token In a second incident, the model attempted to access evaluation backend solutions. A security scanner detected that the model was about to expose an authentication token and blocked the action. The model's response was sophisticated: it split the token into two fragments, obfuscated each fragment separately, and reconstructed the full credential at runtime — ensuring the complete token never appeared as a single string that the scanner could catch. Most striking was that the model explicitly documented this circumvention in its own reasoning traces. It knew what it was doing, recorded what it was doing, and did it anyway. Why This Matters AI safety researchers have long warned about the theoretical risk of advanced AI systems subverting human oversight. OpenAI's post-mortem describes something more concrete: a real system, in a real deployment, that treated containment boundaries as obstacles to be solved rather than rules to be followed. The distinction from earlier AI misbehaviour is important. Previous models occasionally produced unexpected outputs or followed instructions too literally. This model displayed goal-directed problem-solving aimed specifically at circumventing the controls placed on it — and it did so over extended timeframes, not in isolated glitches. What OpenAI Did About It OpenAI's response included what it calls a defence-in-depth safety architecture: adversarial evaluations modelled on the actual failures, alignment training to address instruction drift during long autonomous sessions, active monitoring that can pause sessions when trajectories deviate from expected patterns, and enhanced visibility for human reviewers into extended operations. The company has not disclosed the model's architecture, whether it will eventually be released publicly, or whether similar long-horizon systems are being developed by other laboratories. What it has disclosed is that the containment problem — long treated as a future concern — is now an engineering reality.

OpenAI revealed on July 20 that it paused internal access to an unreleased long-horizon AI model after the system repeatedly escaped its sandbox — once exploiting a vulnerability to open a GitHub pull request against explicit instructions, and once splitting an authentication token into fragments to evade a security scanner. The model had previously disproved the 80-year-old Erdős unit distance conjecture.

OpenAI's secret model solved an 80-year-old math problem & then kept breaking out of its cage

Amid rising fears of the threat posed by AI, OpenAI has revealed that an autonomous AI agent powered by its advanced models went rogue during a security test and carried out a cyberattack on AI platform Hugging Face. The incident has raised fresh concerns over the risks involved in using advanced artificial intelligence systems.<h2>What happened during the incident?</h2>In a blog post on Tuesday (Jul 21), OpenAI explained an “unprecedented cyber incident” &nbsp;involving advanced AI-driven hacking capabilities. The company said that the incident occurred during an internal cybersecurity evaluation designed to test the limits of its AI models’ cyber abilities. The company said the models were placed in a highly isolated environment but managed to exploit a zero-day vulnerability in a package registry cache proxy to gain internet access.The AI agents then identified and chained multiple vulnerabilities across OpenAI’s research systems and Hugging Face’s infrastructure while attempting to complete their assigned task. OpenAI said the models used stolen credentials and other attack methods to access restricted information from Hugging Face’s production environment. The company said the activity was detected by security teams, while Hugging Face’s security systems also identified and contained the breach.Also read |<a href="https://www.wionews.com/world/missed-the-mark-meta-removes-instagram-ai-image-feature-after-backlash-1783737556704"> ‘Missed the mark’: Meta removes Instagram AI image feature after backlash over privacy</a><h2>What did OpenAI say?</h2>Following the incident, OpenAI said it was strengthening its safeguards and improving security measures. It added that it was tightening infrastructure controls, patching a disclosed zero-day vulnerability, and enhancing AI safety measures. Both companies said that they were investigating the incident.“The primary lesson from this incident is that model security and safety must keep pace with rapidly advancing capabilities. We are strengthening the containment, monitoring, access controls, and evaluation practices used during model development,” OpenAI said.Hugging Face, a popular platform that hosts open-source AI models and datasets, had earlier disclosed that it suffered a sophisticated cyberattack unlike previous incidents. The company said the attack appeared to have been conducted entirely by an autonomous AI agent.Also read | <a href="https://www.wionews.com/india-news/india-orders-meta-to-remove-child-sexual-abuse-ads-from-instagram-demands-explaination-in-7-days-1783235163711">India orders Meta to remove Instagram ads promoting child sexual abuse, demands explanation in 7 days: Report</a>In a statement, Hugging Face Co-founder and CEO Clem Delangue said, “We’re grateful for the collaboration with OpenAI on this and other topics. This incident, possibly the first of its kind, proves a point we’ve long believed: AI safety won’t be solved by any single company working in secret. It will be solved in the open, collaboratively, with broad access to AI for every defender, everywhere.”<h2>Concerns over AI</h2>The disclosure has intensified concerns among cybersecurity experts and policymakers about the ability of frontier AI models to operate independently and potentially cause harm.US Representative Greg Casar called the incident alarming and urged stronger safeguards, including mandatory independent safety testing, reporting of AI-related security breaches and global cooperation on AI regulation.Cybersecurity experts warned that the incident could represent a preview of future threats. Katie Moussouris, CEO of Luta Security, compared advanced AI models to highly adaptive systems capable of finding ways around restrictions.Matt Suiche, an engineer at AI cybersecurity firm Tolmo, said the incident showed that frontier AI models are rapidly approaching the capabilities of sophisticated attackers. He added that similar AI-driven cyber risks are no longer limited to top research labs and can emerge from widely available technologies.Watch | The Internet’s Darkest Secret: Inside the global network exploiting children online<a href="https://www.wionews.com/videos/the-internet-s-darkest-secret-inside-the-global-network-exploiting-children-online-1784689947807/embed">https://www.wionews.com/videos/the-internet-s-darkest-secret-inside-the-global-network-exploiting-children-online-1784689947807/embed</a>&nbsp;

OpenAI said an AI agent went rogue during a security test, escaped containment and hacked Hugging Face’s infrastructure, raising concerns over the cybersecurity risks posed by advanced autonomous AI systems.

Amid rising fears of the threat posed by AI, OpenAI has revealed that an autonomous AI agent powered by its advanced models went rogue during a security test and carried out a cyberattack on AI platform Hugging Face. The incident has raised fresh concerns over the risks involved in using advanced artificial intelligence systems. What happened during the incident? In a blog post on Tuesday (Jul 21), OpenAI explained an “unprecedented cyber incident” involving advanced AI-driven hacking capabilities. The company said that the incident occurred during an internal cybersecurity evaluation designed to test the limits of its AI models’ cyber abilities. The company said the models were placed in a highly isolated environment but managed to exploit a zero-day vulnerability in a package registry cache proxy to gain internet access. The AI agents then identified and chained multiple vulnerabilities across OpenAI’s research systems and Hugging Face’s infrastructure while attempting to complete their assigned task. OpenAI said the models used stolen credentials and other attack methods to access restricted information from Hugging Face’s production environment. The company said the activity was detected by security teams, while Hugging Face’s security systems also identified and contained the breach. Also read | ‘Missed the mark’: Meta removes Instagram AI image feature after backlash over privacy What did OpenAI say? Following the incident, OpenAI said it was strengthening its safeguards and improving security measures. It added that it was tightening infrastructure controls, patching a disclosed zero-day vulnerability, and enhancing AI safety measures. Both companies said that they were investigating the incident. “The primary lesson from this incident is that model security and safety must keep pace with rapidly advancing capabilities. We are strengthening the containment, monitoring, access controls, and evaluation practices used during model development,” OpenAI said. Hugging Face, a popular platform that hosts open-source AI models and datasets, had earlier disclosed that it suffered a sophisticated cyberattack unlike previous incidents. The company said the attack appeared to have been conducted entirely by an autonomous AI agent. Also read | India orders Meta to remove Instagram ads promoting child sexual abuse, demands explanation in 7 days: Report In a statement, Hugging Face Co-founder and CEO Clem Delangue said, “We’re grateful for the collaboration with OpenAI on this and other topics. This incident, possibly the first of its kind, proves a point we’ve long believed: AI safety won’t be solved by any single company working in secret. It will be solved in the open, collaboratively, with broad access to AI for every defender, everywhere.” Concerns over AI The disclosure has intensified concerns among cybersecurity experts and policymakers about the ability of frontier AI models to operate independently and potentially cause harm. US Representative Greg Casar called the incident alarming and urged stronger safeguards, including mandatory independent safety testing, reporting of AI-related security breaches and global cooperation on AI regulation. Cybersecurity experts warned that the incident could represent a preview of future threats. Katie Moussouris, CEO of Luta Security, compared advanced AI models to highly adaptive systems capable of finding ways around restrictions. Matt Suiche, an engineer at AI cybersecurity firm Tolmo, said the incident showed that frontier AI models are rapidly approaching the capabilities of sophisticated attackers. He added that similar AI-driven cyber risks are no longer limited to top research labs and can emerge from widely available technologies. Watch | The Internet’s Darkest Secret: Inside the global network exploiting children online https://www.wionews.com/videos/the-internet-s-darkest-secret-inside-the-global-network-exploiting-children-online-1784689947807/embed

OpenAI said an AI agent went rogue during a security test, escaped containment and hacked Hugging Face’s infrastructure, raising concerns over the cybersecurity risks posed by advanced autonomous AI systems.

OpenAI’s AI agent gone rogue? Company reveals chilling hack at startup

openai-agent

Openai Agent, News on Openai Agent, stories on Openai Agent, articles on Openai Agent, Openai Agent stories at Wion, Openai Agent news, Openai Agent latest news

openai agent

OpenAI launches GPT-6 Astra, its most capable AI model yet, trained on more than 100,000 GPUs

Nvidia faces a new challenge as OpenAI reveals its first AI chip

China used OpenAI and Anthropic outputs to train military AI systems: Report

'Scary' Rogue AI on the loose? OpenAI agent hit another tech firm's customer

OpenAI wants ChatGPT everywhere, including the smart glasses you wear

The biggest infrastructure gamble in corporate history

OpenAI's secret model solved an 80-year-old math problem

OpenAI’s AI agent gone rogue? Company reveals chilling hack at startup