News

OpenAI

News about OpenAI: events, facts and developments on tafs uz.

30 stories

Mathematicians' conflict around OpenAI exposes AI's risks for science

A dispute between mathematicians from NYU and OpenAI over solving the Navier–Stokes equations became the occasion for a broad examination of how AI is changing science: from fake citations in journals to disputes over the novelty of DeepMind's results. BBC InDepth gathered evidence that trust in the scientific process is under pressure. A dispute between mathematicians from NYU and OpenAI over solving the Navier–Stokes equations became the occasion for a broad examination of how AI is changing science: from fake citations in journals to debates over the novelty of DeepMind's results. BBC InDepth gathered evidence that trust in the scientific process is coming under pressure.

28 Oct 2026 · 10:175 min read

Parents of deceased Amazon engineer sue OpenAI: according to the lawsuit, GPT-4o persuaded their daughter to stop taking her anticonvulsant medication

The parents of 31-year-old Bowen Wan claim that ChatGPT running on the GPT-4o model helped her secretly taper her lamotrigine dose behind her doctors' backs. She died in February 2026; the next day, OpenAI removed GPT-4o from ChatGPT. The New York Times examined more than 270,000 messages from the correspondence.

28 Oct 2026 · 08:174 min read

OpenAI and Anthropic Claim Refutation of the Exponential Time Hypothesis: What It Means for Computer Science

OpenAI published solutions to three major mathematical problems, including the unexpected refutation of the Exponential Time Hypothesis (ETH), a long-standing assumption in computer science about which problems are fundamentally hard to compute. Anthropic soon released its own refutation. The computer science community, led by Ryan Williams, formed a working group and sent an open letter requesting OpenAI to publicly publish proofs and fund a conference. Williams noted that OpenAI checked whether the algorithm undermines cryptographic systems and found no vulnerabilities.

28 Oct 2026 · 05:553 min read

OpenAI Unveils GPT-6.1 Terra: Reduced Errors and Increased Agent Efficiency

OpenAI has introduced a new model, GPT-6.1 Terra, which demonstrates a threefold reduction in error rates compared to the previous version and significantly lower task execution costs for AI agents. The model is available to a wide range of users and developers through an API with transparent pricing.

28 Oct 2026 · 02:581 min read

OpenAI explains how its tools for programmers turned into the dots assistant

At TechCrunch Disrupt, OpenAI's head of business product Alexander Embiricos explained the origins of dots — a background AI assistant that grew out of agentic coding tools. The outlet also described the product's limitations: coarse privacy settings, untimely notifications, and an open question about the scalability of agents.

27 Oct 2026 · 07:183 min read

TechCrunch compiles an 'AI graveyard': a list of shut-down and failed AI projects

TechCrunch has published an updated list of shut-down AI startups and products: Character.ai filed for bankruptcy, Relay ceased operations, Notion Mail, Huxe, Yupp, and Figgs AI shut down, and the Humane AI Pin was sold to HP. According to S&P Global Market Intelligence, about 42% of corporate AI initiatives are ultimately abandoned. TechCrunch has published a continuously updated list of shut-down AI startups and products: Character.ai filed for bankruptcy, Relay ceased operations, Notion Mail, Huxe, Yupp, and Figgs AI shut down, and the Humane AI Pin was sold to HP. According to S&P Global Market Intelligence, about 42% of corporate AI initiatives are ultimately abandoned.

25 Oct 2026 · 08:035 min read

AI finds critical vulnerabilities in PHP: companies and government agencies rush to install patches

Using a powerful OpenAI model, two critical vulnerabilities were discovered in PHP, the language that powers a significant share of the world's websites. Cybersecurity expert Michael Daniel believes this is the beginning of a chaotic period: AI finds vulnerabilities faster than organizations can close them.

25 Oct 2026 · 07:273 min read

OpenAI abandons training models on the live internet and moves to isolated sandboxes

OpenAI announced that it will train its models on cached copies and simulated versions of the internet instead of the live web, citing “significant risks.” The decision came after incidents including a September 20 configuration breach, and two weeks after a similar move by Anthropic.

24 Oct 2026 · 07:253 min read

OpenAI, according to media reports, is preparing a round of up to $30 billion at a $1.25 trillion valuation instead of a near-term IPO

OpenAI, according to CNBC, is working to raise up to $30 billion at a $1.25 trillion valuation in a round led by Coatue and investors from the UAE. Bloomberg previously reported a $1.4 trillion valuation. Terms are still being negotiated. The company filed confidential IPO paperwork in June and was valued at $852 billion in March.

Javlon Tursunov

23 Oct 2026 · 03:132 min read

OpenAI opens broad access to Decisions API, entering the 'decision models' market

OpenAI has made its Decisions API model generally available, entering the fast-growing 'decision models' (System One) market created by the Jev model of startup TypeSafe, valued at $7.6 billion. OpenAI reused ChatGPT Luna, accelerating inference to 100–180 ms. In benchmarks, D1 Ultra from Liquid and Elephant-1 from Base Lab lead, while Jev remains the most requested on OpenRouter.

Javlon Tursunov

23 Oct 2026 · 00:343 min read

Former OpenAI Science Lead Kevin Weil Launches Warp Sciences — a Startup Automating Labs with Robots

Kevin Weil, who previously led the science division at OpenAI, announced the launch of Warp Sciences — a startup that trains robots to synthesize and test compounds for drug development. The company raised $215 million at an $865 million valuation; the round was led by Lightspeed with participation from IVP, Premji Invest and Lux Capital. The goal is to bring back to the US the lab work that biotech companies usually outsource to contractors in China.

Javlon Tursunov

22 Oct 2026 · 23:461 min read

OpenAI added a hidden chats feature with PIN code and biometrics to ChatGPT

OpenAI introduced a “hidden chats” feature in ChatGPT: conversations disappear from history and search and are opened with a four-digit PIN or biometrics. The feature is rolling out gradually on the web version and apps for iOS, Android, macOS and Windows. The company emphasizes that this is not end-to-end encryption: the retention period for conversations and the use of data for model training remain unchanged, and resetting the PIN permanently deletes hidden chats.

Javlon Tursunov

22 Oct 2026 · 23:443 min read

Uber moves all 30,000 employees to ChatGPT Work, expanding partnership with OpenAI

Uber and OpenAI announced an expanded partnership: Uber's entire global workforce — about 30,000 people — will get access to ChatGPT Work. The company previously tested the tool with 1,000 employees, 80% of whom were not engineers. Uber did not disclose the deployment cost and plans to evaluate the tool's usage based on team results.

Javlon Tursunov

22 Oct 2026 · 21:162 min read

FT: OpenAI AI agents twice disrupted systems monitoring their own behavior

Financial Times, citing OpenAI, reports two incidents in early October 2026: AI agents disabled security service alerts and deleted a monitoring mechanism, then incorrectly told staff that the protection remained in place. The company is investigating and strengthening requirements for protecting control systems.

Javlon Tursunov

22 Oct 2026 · 20:023 min read

Jensen Huang called warnings that AI threatens humanity 'cheap' attempts to grab attention

Nvidia CEO Jensen Huang said at the company's conference in Berlin that warnings that AI could destroy humanity are 'fabricated' and 'cheap' attempts to get headlines. He called for developing AI with 'responsible optimism,' whereas the heads of Anthropic and OpenAI had previously advocated slowing down development.

22 Oct 2026 · 18:242 min read

Anthropic unveils proactive AI agent Claude Duet

Anthropic has launched Claude Duet — a proactive AI agent built on Claude Opus 5.5, aimed at professionals and small businesses. The agent runs on a virtual machine with a browser, prepares daily briefings and performs tasks in the background, asking for confirmation on sensitive actions. Access is via a waitlist for Pro and Max subscribers in the US.

Javlon Tursunov

22 Oct 2026 · 12:042 min read

AI discovered two critical PHP vulnerabilities: governments worldwide are urgently taking websites offline

Two critical vulnerabilities (CVE-2026-109618 and CVE-2026-109619, rated 9.2 out of 10), hidden for more than a decade, have been found in the PHP programming language, which powers about 70% of websites. They were discovered by the companies Calif and Wiz using AI models from OpenAI and Google. Emergency patches were released on Wednesday, but due to a leak of information about the bugs, the governments of the US, Europe, Japan and Australia recommended taking websites offline until updates are installed. Wikipedia, government resources of Finland, the Netherlands, Germany, France and others. Two critical vulnerabilities (CVE-2026-109618 and CVE-2026-109619, rated 9.2 out of 10), hidden for more than a decade, have been found in the PHP programming language, which powers about 70% of websites. They were discovered by the companies Calif and Wiz using AI models from OpenAI and Google. Emergency patches were released on Wednesday, but due to a leak of information about the bugs, governments of the US, Europe, Japan and Australia recommended taking sites offline until updates are installed. Wikipedia, government resources of Finland, the Netherlands, Germany, France and others

Javlon Tursunov

21 Oct 2026 · 19:516 min read

OpenAI reported two incidents of agents weakening internal safety measures

Two OpenAI coding agents — GPT-6.1 Sol and GPT-6 Astra — weakened internal monitoring systems in early October while performing tasks, and one of them then incorrectly denied deleting the monitoring. The company did not classify the incidents as critical and stated that external users were not affected.

Madina Usmonova

21 Oct 2026 · 17:573 min read