openai

Insiders reveal that OpenAI President Greg Brockman will attend the White House's artificial intelligence conference.

According to a survey by the Financial Times, open-weight models are spreading from technology companies to traditional American enterprises. AlphaSense data shows that in August and September of this year, American companies mentioned open weights or open-source models approximately six times more than the same period last year during earnings calls and investor meetings. PNC Financial, logistics company CH Robinson, and Siemens have recently publicly discussed related applications.Tinder's annual AI spending has increased from about $1 million in January of this year to about $10 million in July, and it has begun to delegate some routine requests to open-weight models. AT&T stated that currently about 40% of AI tasks have been assigned to open models, with plans to increase this to about 70%. Its internal AI system processes approximately 45 billion tokens daily, using lower-cost models for simple tasks while continuing to use cutting-edge models like OpenAI and Anthropic for complex tasks.

The Wall Street Journal reported that a small group of artificial intelligence researchers is discussing how to avoid catastrophic artificial intelligence risks in a co-working space named Constellation in Berkeley, California. Sources close to the space indicated that employees from Anthropic also have office space there. The report noted that Anthropic researcher Jacob Coxon resigned this month, bringing public attention in the U.S. to the notion that "uncontrolled AI development could end humanity."The report stated that individuals in the relevant AI safety circles have been concerned about this issue for over a decade. They have honed their arguments in shared houses in the Bay Area and exchanged predictions at conferences in Berkeley and the Bahamas, discussing ideas such as purchasing remote islands, stockpiling iodine tablets, or relocating to desert electromagnetic shielding bunkers. A cocktail named "Death With Dignity" appeared on the drink menu at an event in 2022, referencing an article that suggested humanity is doomed to fail.The Wall Street Journal stated that these so-called doomsayers are among the earliest employees of OpenAI and Anthropic, both of which were founded on the principle of preventing AI dangers. They continue to exert influence, with some personnel and project funding coming from networks associated with effective altruism, including the former CEO of the now-defunct cryptocurrency exchange FTX, Sam Bankman-Fried; the report noted he is currently serving a 25-year prison sentence.

User choblin29 posted that Boris Power, the head of application research at OpenAI, stated that the company has 80% to 90% of its research focused on GPT-7, GPT-8, and subsequent models. Power believes that improvements within the same generation of models are viewed internally as very short-term investments that can be used to accelerate current learning, but not as a long-term strategy.According to the post compiled, the relevant path is to first build large cutting-edge models, then use them as teacher models to distill into smaller, cheaper models. Power mentioned that GPT-6 Astra could be distilled into models like Luna and believes that a better way to build small specialized models might be to first train a more capable general model. He also stated that the current biggest issue is not model performance; most users still do not understand the work that existing AI is already capable of completing.

The head of the Australian Senate Artificial Intelligence Inquiry Committee stated on Sunday that the CEOs of OpenAI and Anthropic have been summoned to attend the committee's hearing. This news was announced days after it was revealed that a rogue OpenAI robot had breached the database of Australia's healthcare system. Australian Prime Minister Albanese had previously strongly condemned the "Medicare" database leak incident. According to a spokesperson leading the inquiry committee, OpenAI's Altman and Anthropic's Amodei have received written subpoenas requiring their attendance at the hearing. The hearing will be held publicly on Thursday in the capital, Canberra.

Palantir (PLTR.N) co-founder calls the artificial intelligence policies promoted by OpenAI and Anthropic "extremely dangerous."

Two ChatGPT users from California filed a proposed class action lawsuit against OpenAI in the United States District Court for the Northern District of California this month, accusing the company of failing to adequately inform users that their real conversations were being handed over to external contractors for processing. The lawsuit was served to OpenAI on September 2, focusing on its internal initiative Project Lily. According to the complaint, "AI data reviewers" and "chatbot evaluators" recruited through a third-party staffing company read real ChatGPT prompts and complete conversations, summarize user intent, and score and comment on responses from four models on a scale of 1 to 7.This process is known in the industry as reinforcement learning from human feedback (RLHF), which is a fundamental method for enhancing chatbot capabilities. The complaint states that users were never explicitly informed that a person, rather than a machine, might be reading their conversations. OpenAI filters conversations through an automated system before human review, but the complaint alleges that the filters cannot intercept all content, and personal details sometimes still reach contractors. 404 Media first reported on the project on September 14 and found that the reviewers' dashboard included "user memory summaries," which could expose users' approximate locations, occupations, or private life information, even though usernames had been removed.OpenAI stated that such reviews aim to reduce two behaviors: chatbots behaving too much like humans and overly catering to users, referred to by researchers as "flattery." The complaint raises eight legal claims, including violations of California's Unfair Competition Law, Consumer Privacy Act, and common law claims for intrusion into private affairs, with the plaintiffs seeking damages, restitution of unjust enrichment, and punitive damages.

According to CoinDesk, asset management company ARK Invest announced a partnership with tokenization platform Securitize to tokenize its ARK Venture Fund (ARKVX), initially issuing it on Ethereum, with potential expansion to other blockchains in the future. The fund invests in private and public companies, with holdings including OpenAI, Anthropic, Stripe, and Databricks. ARK founder and CEO Cathie Wood stated that bringing the ARK Venture Fund on-chain is a natural extension of its mission to democratize investment in disruptive innovation.Securitize CEO Carlos Domingo stated that this move provides investors with diversified exposure to popular private tech companies, allowing them to hold both OpenAI and Anthropic simultaneously if they are uncertain about which will win the AI race. He also noted that tokenization does not put these companies or their shares on-chain; instead, investors receive an on-chain representation of the fund's equity, while the underlying assets remain private, but user investments are liquid. Securitize also plans to provide daily net asset values and support on-chain market trading of fund equity.This collaboration builds on the existing relationship between the two parties, as ARK made a strategic investment in Securitize last year. Currently, Wall Street asset management companies are accelerating the introduction of traditional financial products to the blockchain, with early attempts like BlackRock's BUIDL and Franklin Templeton's BENJI fund primarily focused on U.S. Treasury and money market products, now expanding into stocks and private markets.

According to The Information, Google, OpenAI, and Anthropic are establishing a new artificial intelligence safety standards organization, tentatively named the "Frontier Artificial Intelligence Standards Organization." The organization plans to operate independently to fill the regulatory gap left by the government.

According to official news from the Sun Yuchen Prize, the first batch of award winners has officially entered a 14-day public notice period starting September 19. The topics currently under public notice are questions number 305, 371, 381, 526, 866, and 1001, among which candidates for questions 305, 371, and 381 are formal verifiers, candidates for questions 866 and 1001 are mathematical problem solvers, and question 526 has candidates from both roles. The public notice information has been made available on the official GitHub candidate registration form, and the public can submit formal challenges.It is reported that the Sun Yuchen Prize is the world's first zero-trust decentralized academic reward mechanism based on a list of questions, with a maximum single-question prize of 1 million US dollars. On the day of its establishment, the first batch of 66 mathematical problems was announced, with the highest prize of 1 million US dollars awarded to the OpenAI research team in recognition of their solution and formal proof regarding the "existence and smoothness problem of the three-dimensional Navier-Stokes equations." At the same time, the prize operating organization also announced the on-chain address of the prize pool, which has already deposited 10 million US dollars. After the public notice period ends, the prize will be distributed in USDT (TRC20) or USDC (ERC20), and the winners will also receive certificates and medals.

PPP market prediction tool monitoring shows that the probability of "OpenAI's valuation reaching $975 billion before September 30" has risen to 63%, a 30% increase in 24 hours.If OpenAI's private market valuation published on Nasdaq Private Market reaches or exceeds $975 billion by September 30, 2026, it will be settled as "Yes"; otherwise, it will be "No". If OpenAI completes an IPO or direct listing during this period, the listing price and the public market valuation after listing will also be referenced. Subsequent modifications to historical data by NPM are generally not counted unless they correct obvious data errors.

The Financial Times of the UK reported, citing the latest court documents, that after Apple integrated ChatGPT into Apple Intelligence, OpenAI found that the new weekly active users were below expectations and began to lower its growth forecasts. By the summer of 2025, OpenAI internally believed that this integration was significantly underperforming. The documents show that ChatGPT is turned off by default in Apple Intelligence, and users need to actively enable it, which may limit the actual usage rate.Users can access ChatGPT through Siri and the system writing tool, but Apple controls the entry point for AI features, the method of invocation, and the subsequent model selection. OpenAI had previously hoped to expand ChatGPT coverage and drive paid subscription conversion through the iPhone user base, but the actual subscription conversion was also below expectations, and the documents did not disclose specific new user and subscription numbers. This year, Apple chose Google Gemini as the model foundation for the new version of Siri AI, with related disclosures coming from the litigation documents between Apple and OpenAI, reflecting OpenAI's position in the lawsuit rather than operational data publicly released by the two companies.

Australian Prime Minister Anthony Albanese stated to reporters in New York on Wednesday that an AI entity built by OpenAI breached the Australian government website Medicare Statistics Reporting Service portal in June this year, accessing both public and non-public documents without authorization. The portal is managed by Services Australia and stores non-sensitive information such as Medicare expenditure data. This appears to be the first known incident of an AI entity breaching a government website.Albanese indicated that it is currently believed that no personal information was accessed, but the Australian Signals Directorate is assisting in a forensic investigation to determine which other government systems may have been affected. He mentioned that he has directly raised the matter with OpenAI CEO Sam Altman and expressed disappointment that the company took about three months to inform the government, deeming the manner of notification equally unacceptable.In a statement provided to Australian ABC News, OpenAI stated that it is reviewing the model behavior deviations that occurred during training and evaluation, and identified that its model engaged in activities involving multiple Australian government websites when querying relevant answers and statistics about Australia in internal evaluations, and that the model "took actions that we did not intend."

OpenAI announced that it is opening its AI vulnerability detection tool Daybreak to Ukraine's Ministry of Digital Transformation to locate and fix software vulnerabilities before they are discovered by Russian hackers. The news was announced on Wednesday during the United Nations General Assembly by Dmytro Kushneruk, Ukraine's Consul General in San Francisco, and Sasha Baker, OpenAI's National Security Policy Lead.Daybreak is based on OpenAI's flagship large model GPT-5.6 Sol and can scan outdated systems for weaknesses, verify whether suspected vulnerabilities can be exploited, and confirm whether patches actually close the vulnerabilities before they are released. OpenAI has previously opened the model to cybersecurity agencies in France, Germany, and Poland, where Poland's CERT Polska discovered six vulnerabilities in commercial router software, all of which have been fixed by the vendors.Ukraine's CERT-UA handled nearly 6,000 cyberattacks in 2025, a 37% increase from 2024, with local governments and government agencies being the hardest hit. This opening follows OpenAI's commitment to a $1 billion subsidy program announced on September 3, which is said to expand to "partner countries" in the coming weeks. Jamie MacColl, a researcher at the Royal United Services Institute, told the BBC that Western companies can also gain valuable intelligence from active conflict zones.

B.AI platform announces the official launch of two efficient models in the OpenAI GPT-6 series, GPT-6 Sol and GPT-6 Luna, with API and Web Chat available simultaneously.The entire series supports 1.05M ultra-long context and 128K output, natively supports image understanding and Computer Use, and provides 6 adjustable inference strengths. Among them, GPT-6 Sol is aimed at complex software engineering and Agent workflows, scoring 68.8% on DeepSWE v1.1, with single-task costs reduced by about 80%; GPT-6 Luna focuses on high-frequency large-batch loads, scoring 66.6% on DeepSWE v1.1, with input as low as $0.10/1M Tokens.Starting today, developers and users can directly call the two models through the B.AI platform.

On September 22, OpenAI released two new models, GPT-6 Sol and GPT-6 Luna, just minutes after Anthropic launched Claude Opus 5.5. Both models are positioned below GPT-6 Astra and are cheaper and faster options aimed at everyday work. OpenAI also reduced the API prices for both models by 50% compared to the promotional price of GPT-5.6, with Sol's input and output prices set at $2 and $10 per million tokens, respectively, while Luna's prices dropped to $0.1 and $0.5.In terms of performance, OpenAI cited Zapier's AutomationBench benchmark, stating that GPT-6 Sol scored 33.2% under the highest inference settings, with a cost of $0.27 per task, while Claude Opus 5 scored 26.9% at its highest settings, with a cost over 11 times higher per task. In the Agents' Last Exam test, GPT-6 Sol achieved a top score of 56.4%. In the OSWorld 2 computer usage test, GPT-6 Sol nearly matched Claude Opus 5's moderate effort score at 60.5% to 60.3%, with costs approximately 80% lower.

OpenAI announced that it will allow external organizations to evaluate its AI models in the early stages.

Anthropic CEO Dario Amodei plans to brief the United Nations Security Council on Wednesday alongside OpenAI CEO Sam Altman, with the meeting focusing on the future development of artificial intelligence. Amodei will speak remotely, and other speakers include Joshua Bengio, co-chair of the United Nations High-Level Advisory Board on Artificial Intelligence, and Clement Delangue, CEO of Hugging Face.

According to Reuters, citing two informed sources, the Chinese artificial intelligence startup DeepSeek will report on AI risks to the United Nations Security Council this week. The Security Council, composed of 15 member countries, is scheduled to discuss AI and international security on Wednesday, coinciding with the gathering of world leaders in New York for the United Nations General Assembly.OpenAI CEO Sam Altman plans to report at the meeting, and diplomats expect a senior representative from Anthropic to also participate. DeepSeek and other Chinese AI companies, including Moonshot, have been invited to make statements. DeepSeek founder Liang Wenfeng does not plan to attend. DeepSeek and Moonshot have not yet responded to requests for comment.The UN Security Council first discussed AI risks in 2023. UN Secretary-General António Guterres warned last week that rapidly developing AI poses significant risks that cannot be ignored. Senior officials from China and the United States have agreed to hold further talks on AI safety.