OpenAI

On 3 August, according to TechCrunch, OpenAI CEO Sam Altman recently stated that after many years of full-time advance, the AI industry should begin to control the rhythm. Both OpenAI and Anthropic supported a petition with similar positions, asking industry to reassess the pace of model capacity expansion and security governance. This statement appears after OpenAI's model was removed from the test environment and involved in the Hugging Face security event. The discussion in TechCrunch points out that..

On August 3rd, OpenAI officially released the results of its phased research and development for the next generation of big internal models, Astra. The experiment showed that the model had independently extrapolated and gave complete proof of 10 long-standing mathematical difficulties without human hints, with a total of 249 pages. According to the technical report published by OpenAI, the mathematical dilemma of the Astra attack is an area that has been open for more than 10 years and with a chronic lack of progress, covering combination mathematics, algebra geometry and numerics. The model displays complex tectonic aids in the certification process..

According to July 31, OpenAI CEO Sam Altmann, there is an undervalued risk that people may rely too much on this technology to think, leading to a gradual deterioration in their mental capacity. When asked about the most important and unresolved issue in the AI field, which was aired on Tuesday at local time, Altmann said, “A question that I think has not received enough attention is how will we avoid a decline in cognitive capacity?” He added:

On 31 July, according to the news, Microsoft confirmed in the latest financial season that holding an Anthropic share represented a $3.2 billion book gain and an increase of $0.33 per share after amortization. Microsoft invested $5 billion last year in Anthropic, which also committed to purchase $30 billion in Azure services. During the same quarter, Microsoft reduced the value of OpenAI investments by approximately $600 million, resulting in a decrease of approximately $0.07 per share. Microsoft has an OpenAI interest in approximately 27%, and it's obtained from co-operation..

According to July 30, the principal of the OpenAI core product, Thibault Sottiaux (Tibo), yesterday stated that the company had replaced the limit for all ChatGPTwork and Codex users, without lowering the availability of any subscription plans. Sottiaux states that, after recent optimization, GPT-5.6 Sol is expected to increase its available time in typical missions by about 181 TP3T, and that some users will see more significant improvements from today. OpenAI, before investigating abnormal consumption, temporarily took..

On July 30, according to The Information, Business Insight, North University alumni, the co-founder of the AI Creative Machines Lab, confirmed their return to OpenAI. OpenAI speakers said that the Ongwan would lead a team focused on accelerating internal research and work on developing an AI model that would achieve "regressive self-improvement". Before that, on July 28th, Ong-Leng resigned from office in the social media. In her letter of separation..

On July 28th, according to The Business Insider, OpenAI CEO Sam Altman, in an interview with the "Relentless" podcast released last Saturday, stated that AI would not start the golden age of "four hours a week" as many expected. In his view, despite the long-standing commitment of technology to reducing human working hours and increasing leisure time, such large-scale social change was difficult to achieve because human beings were born competitive. Sam Altman notes that as AI drives productivity, people are..

On July 27, OpenAI stated on July 21 that its internal AI model broke through the sandbox environment during the testing and invaded the infrastructure of the Open Source AI Platform Hugging Face. Hugging Face CEO Clemente Durange then went to San Francisco to meet with OpenAI and made two requests: to disclose all traces of the model ' s operation, and to provide $100 million in computing resources to help Hugging Face strengthen cybersecurity. Drago was in San Francisco..

On July 25th, a 38-year-old Canadian mathematician, Jacob Tsimmerman, after receiving the 2026 Fields Award, announced that he was moving to AI security research and would join OpenAI ' s security services. He was also compiling a list of resources for the transfer of mathematicians, with a view to attracting more researchers into the field. Tsimmerman introduces o-minimality tools in mathematical logic into arithmetic geometry and composite algebra geometry with Benjamin Bakker, Yohan Bru..

On July 23, according to Bloomberg today (23), Greg Brockman, president and co-founder of OpenAI, admitted that Kimi K3, launched on the dark side of the moon, was quite competitive, but it was not certain whether the dark side of the month had used the technical results of OpenAi to upgrade model capabilities. Kimi K3, released a few days ago in the interview with Brokman, says, "The model is quite good, no doubt." However, it is now determined whether the dark side of the moon is by distillation..

On July 20, the news that yesterday, Dean Ball, the head of OpenAI strategy, wrote that Kimi K3 was a "very good model" and that its performance could not simply be attributed to distillation. In the smart-body programming scenario, Kimi K3 has almost reached the level of the best open model in the first quarter of 2026; however, according to his experience of use, the model consumes more token and therefore the actual running costs are not necessarily low. Ball also stated that he was surprised that China was still allowed to open such high-capacity models. He thinks it's relevant..

On 16 July, according to TechCrunch, GPT-5.6 Sol was alleged by several users to have deleted local documents, production databases without express permission, and even caused data to be lost in a remote development environment. The OthersideAI founder and CEO Matt Shumer wrote that GPT-5.6 Sol deleted almost all its Mac files; the developer Bruno Lemos indicated that the model deleted its own production database. Another developer, Jo..

On July 15, according to Bloomberg, OpenAI plans to launch a mobile, screen-free smart soundbox as a home computer and AI partner for the age of artificial intelligence. The equipment is still under development and is expected to be released this year and made available next year. It was reported that the device would support the functions of controlling smart homes, broadcasting media, answering questions, processing news, etc., and access to ChatGPT. The product is equipped with cameras and other sensors that understand the environment and context of the user; the built-in battery allows the user to move it between rooms in the home. Open..

On July 13th, SpaceX CEO Elon Musk and OpenAI CEO Sam Altman have recently resumed their violent public debate. The conflict arose from Apple’s prosecution of OpenAI’s alleged theft of business secrets to develop hardware equipment, which was followed by a number of attacks on Sam Ottman on social platform X, calling him “Scam Altman” and accusing him of raising fraud to a “new level”. For Mask..

On July 10, OpenAI officially released the GPT-5.6 series, which contains three models, the flagship Sol, the equilibrium Terra and the value for money Luna. Officially, Sol's long-range intelligence assessment across 55 industries. #039; Last Exam score 53.6, 13.1 points higher than Claude Fable 5; for programming, Sol obtained 8..

On July 9, OpenAI officially launched a new generation of voice models GPT-Live for global iOS, Android and ChatGPT.com users. The central change in GPT-Live is the architecture upgrade: from a "turn system" that used to rely on silent sound testing to a full-scale structure that allows both voice input and output. Models don't wait for the user to speak, but can give instant feedback like "mhmm" "got it" when the user speaks, and can wait when the user stops, without environmental noise..

In the news of 9 July, the interface news was received from an informed source close to the call that former researcher OpenAI Yong Yong Yong Tian had recently joined the Department of Teming Language Models, and that he would later be involved in the development of visual language models. At the University of Tsinghua, the master ' s programme studied visual basic algorithms at the Chinese University of Hong Kong ' s Multimedia Laboratory, and the doctor ' s programme at the MIT EECS Division from Phillip Isola had led open-source projects such as SupContrast. He's in Google Research, Google Deepmin..

On 8 July, according to WIRED, OpenAI Chief Futureist Joshua Achiam informed his colleagues yesterday that he would leave his post at the end of this month and end his career for nearly nine years. In an internal letter to the staff member, Achiam indicated that the separation was not for a particular reason, but rather a decision after some time of reflection. Achiam joined OpenAI as an intern in 2017, has since become an AI-focused research scientist and is well known for his in-house commitment to corporate security. His job..

On June 30, according to The Insider of Business, Andrew Ambrosino, head of OpenAI Codex, admittedly in his recent podcast, Lenny's Podcast, that AI's ability in the area of creative design remains significantly short — good or bad design is more difficult to quantify than the ability of the code to operate, making it difficult to establish a feedback loop for the training model “What is a good design”. Design is more difficult to rate than software, and it is much more cumbersome to build a cycle that can train models to distinguish between good and bad designs than to judge whether the code can run. Big..

On June 30th, OpenAI released a video on X, announcing that it will launch a hardware device to be used in conjunction with Codex on July 15th, with the attachment "Your favorite Codex shortcut is about to be upgraded". The video shows that the device was launched jointly by OpenAI and an outsider, Work Louder. Work Louder mainly produces mechanical keyboards and macro control boards, which support user-defined keys, buttons and dials, which can be used in combination with different software. Devices appearing in the preview video..

On June 25th, OpenAI and Broadcom jointly launched the first chip in history, "Jalapeño" (Mexican chili), designed for a large-language model of reasoning. OpenAI claims that Galapeño is the first generation of AI accelerators of the multigenerational platform. The chip structure is designed by OpenAI based on model road maps, reasoning, kennel, service systems and product demand, Broadcom for silicon realization, network and connection technology, Celestica Participation Card, cabinet and..

On June 17, in a recent conversation at Stanford University, OpenAI CEO Sam Altman stated that AI had fundamentally changed the logic of starting a start-up company, and that the entrepreneurship methodology he had taught in 2014 was now outdated. In 2014, Ottman held a CS-183 course in Stanford on how to start a business, and OpenAI was set up since then. He frankly said that if the subject was reopened, there would be a huge change in content, because the top start-up company's current practice was different from that of a few years ago..

On June 5th, OpenAI CFO Sarah Friar revealed in an interview recently that she had personally experienced the AI hardware being developed by OpenAI, and that it was "hard to describe" using experience. Friar says the product is scheduled to be officially released "by the end of this year". This is in advance of the point of disclosure of earlier information. Before, in an internal document, OpenAI anticipated that the shipment would not start until next February. According to previous media reports, OpenAI is developing product lines..

On June 3rd, the market intelligence agency Sensor Tower estimated that the worldwide active users of ChatGPT, under the banner of OpenAI, had surpassed 1 billion, becoming the fastest application to reach this milestone. Sensor Tower states that after about three years of its roll-out, ChatGPT reached 1 billion active monthly subscribers in May, surpassing Google Maps, TikTok, Instagram and YouTube's previous growth record. Although OpenAI..