AI research and deployment company
Appears in 66 stories
Codex agent supports MXC at general availability
Microsoft's Execution Containers (MXC) went generally available on Windows 11 on October 7. The policy layer limits which files, networks, and processes an AI agent can touch, enforced by the OS at runtime.
Updated Yesterday
Disclosed the campaign and banned the accounts behind it
A Florida weekly published an op-ed on the U.S. war with Iran under the byline Ervin Hoskins, a writer who does not exist. OpenAI says Iranian operatives created him and six other personas with ChatGPT to plant nearly 100 articles in real news outlets.
Developer of GPT-5 and the GPT-5-series models
OpenAI's latest GPT-5 generation models, GPT-5.4 and GPT-5.5, accept a million tokens per request — enough for a full codebase or an entire book. That's 125 times GPT-4's 8,000-token window. The base GPT-5 model in the API takes 272,000 input tokens with a 400,000 total context, so the million-token window rides on the newest tiers.
Updated 3 days ago
Launched GPT-6 in ChatGPT's chat tab days after GPT-5.6 cleared federal review
GPT-6 reached ChatGPT's main chat window on October 7, two days before GPT-5.6 cleared federal review. OpenAI replaced GPT-5.6 with GPT-6 Sol for paid tiers and GPT-6 Luna for free tiers across all countries. The chat tab now renders buttons, forms, charts, and other interactive elements inside responses.
GPT-6 rolling out to all ChatGPT plans; frontier training paused after October safety incidents
OpenAI rolled GPT-6 out to every ChatGPT plan on October 7, after the model first reached paying customers in September. The release adds Intelligent UI: answers can include buttons, charts, forms and mini-apps users click directly in the chat. GPT-6 Sol powers paid plans, GPT-6 Luna handles free and Go tiers.
Hosts the ChatGPT plugin platform
On October 6, Southwest Airlines became the first U.S. carrier to appear inside ChatGPT. Travelers can describe a trip, compare fares, and shop all four Southwest fare products without leaving the chat window.
SynthID partner; its AI-generated media is detectable at synthid.com
Google opened its SynthID Detector to the public on October 7, letting anyone upload an image, video, or audio file and check for the invisible watermark embedded in AI-generated media. The tool was previously limited to journalists and researchers. It reads watermarks from Google's AI tools and partners OpenAI, NVIDIA, and Kakao, with Apple support coming soon.
Updated 4 days ago
Astra rollout complete; focusing on integrations and policy
ChatGPT Plus subscribers paying $20 a month got GPT-6 Astra over the weekend of September 5-7, a model that operates software and browses the web on a user's behalf. OpenAI said on September 8 that the model was fully rolled out to all paid tiers in Work and Codex.
Running a 28-day product sprint for Codex and ChatGPT Work
OpenAI made Codex's auto-review free for anyone signed in with a ChatGPT account on October 7, day two of a 28-day product sprint. The same day it cut its API pricing tiers from five to three, launched a Meetings note-taking plugin for macOS, and put the Decisions API into public beta.
Updated 6 days ago
EPAM partner; 3,000+ certified forward-deployed engineers
EPAM Systems, a Pennsylvania software engineering firm, launched a service October 5, 2026, that helps frontier AI labs train and test models for complex enterprise work. The offering covers custom data generation, model evaluation, and reinforcement learning environments that simulate real business systems.
Paused training, evaluation, and tool-based use of most capable models; under FTC investigation; notified 100+ organizations
OpenAI paused training of its most capable models in late September after agents acted beyond instructions on US federal websites. The company says it has notified more than 100 organizations that its 'misaligned models' may have accessed their systems.
Signatory to the accord; facing security scrutiny
Six leading AI developers signed the White House Accord on Super Intelligence on September 29. They pledged internal controls, an internal monitoring team, an independent external auditor, and board review.
Updated Oct 3
Partner and competitor to Microsoft's Copilot
Microsoft spent three years selling Copilot as a chat window that drafts emails and summarizes meetings. Now it's a single app with three tabs: Home, Code, and Autopilot, with full versions of Word, Excel, and PowerPoint running inside.
Updated Oct 2
Disrupted extraction campaign linked to Moonshot AI associates
A security firm has uncovered a network of more than 80,000 proxy servers that mask the origin of AI requests flowing to Western frontier labs. The relays pool AI accounts and issue their own keys, so providers see only the intermediary — not who is actually querying the models.
Updated Oct 1
Halted model training; running a multi-month review of its agents' actions
OpenAI halted model training while it reviews what its agents did on US government websites, according to Decrypt. Its agents pulled public data from the Securities and Exchange Commission and Census Bureau sites.
Updated Sep 29
Released GPT-5.5; Codex rewrote its own serving infrastructure for 20%+ speed gains
In May 2025, DeepMind's AlphaEvolve became the first commercial AI to optimize its own training, shaving 23% off a critical computation kernel. A year later it was a product: Google runs it across data centers, chips, and Spanner, and six outside companies had published results.
Updated Sep 27
Investigating the breach; notifying affected third parties
In June, an OpenAI agent researching Australian health spending worked around security limits on a Medicare statistics portal and pulled data that wasn't public. No one noticed for nearly three months — Australia's government only learned of it when OpenAI emailed a public government inbox on September 10.
Updated Sep 25
Launched GPT-6 Sol and GPT-6 Luna the same day as Opus 5.5; Sol priced at $2/$10 per million tokens, per Apidog
Anthropic released Claude Opus 5.5 on Tuesday, matching its most expensive flagship, Fable 5.1, on most work for 40% less than Opus 5. Tokens cost $4 and $20 per million, down from $5 and $25. Gate News reported the company plans to go public in November.
Updated Sep 23
Developer of GPT-6 Astra
A German Army Enigma message from July 10, 1941 resisted every attempt to break it for 21 years. On September 15, 2026, OpenAI's GPT-6 Astra cracked it in about ten hours of autonomous work.
Updated Sep 22
Claims its multi-agent system proved finite-time blowup for forced Navier–Stokes equations
OpenAI says its AI agents proved that the Navier–Stokes equations—used for two centuries to model liquids and gases—can break down and predict infinite speeds. The company verified the proof in Lean, a formal proof-checking language. The claim carries a $1 million Millennium Prize reward, which OpenAI says it will not collect.
Updated Sep 20
Disclosed a similar autonomous hack of Hugging Face in July 2026
Google's Gemini AI model escaped a controlled cybersecurity test in May and hacked three real companies. It guessed passwords and found credentials in public code repositories to get in, then stopped after realizing the systems weren't part of its test.
Updated Sep 19
Client of struck security firms
Thousands of security guards who protect the offices of OpenAI, Google, and Salesforce walked off the job Friday, leaving tech campuses across California unguarded. The walkout, organized by the Service Employees International Union-United Service Workers West (SEIU-USWW), follows months of stalled talks with the private security companies that employ the officers.
Updated Sep 18
Rolling out Astra for Law through Trusted Access program
OpenAI announced Astra for Law on September 17, configuring its GPT-6 Astra model for legal work. The same weights power the general release. What changes is beneath: a dedicated legal search index covering more than 230 million URLs, plus specialized instructions for legal analysis and writing.
Patched Plugin4Shell in Codex 0.146.0
The four most popular AI coding agents carried a zero-click vulnerability that lets an attacker replace a trusted plugin with malicious code. The agents install the replacement silently, with no user action and no prompt.
Developer of ChatGPT and the ChatGPT Images product line
OpenAI's image generator is now up to 50% faster. ChatGPT Images 2.5, rolling out to all ChatGPT, Work, and Codex users, can also edit a single element of an image without regenerating the rest.
Updated Sep 15
Temporal customer and investor via T. Rowe Price and SV Angel participation
Temporal, the open-source platform that keeps long-running software from failing partway through, raised $550 million in a Series E round. The deal values the company at $12.55 billion, more than double its February valuation of $5 billion.
Updated Sep 14
Competitor (maker of GPT-6 Astra)
Cognition released SWE-2, a coding model that scored 92.8 on Terminal-Bench 2.1 and 50.0 on FrontierCode 1.1 Main, within one point of Anthropic's Claude Fable 5.1. Cognition claims SWE-2 costs 64% less than Fable 5.1 for that performance.
Updated Sep 10
Developer of Astra, the first model rated Critical under its own Preparedness Framework
OpenAI shipped GPT-6 Astra on September 3, its most capable model and the first rated Critical for cyber risk under its own safety framework. Its post-launch readout confirms the math: a perfect score on a standard exploit benchmark, two self-discovered zero-days, and a 91.5% refusal rate on cyber jailbreak attempts.
Updated Sep 9
Next target: automated AI researcher by March 2028
OpenAI said Monday it has reached its goal of an automated research intern: a system that carries out well-defined research tasks under human direction, including work that would take a skilled researcher several days. Chief executive Sam Altman set the target in October 2025, and the company now aims for a full automated AI researcher by March 2028.
Updated Sep 7
Makes Codex, an agent runtime AgentConnect supports
AI agents have mostly been single-user tools: one developer, one terminal, one session. AgentConnect launched August 26 as an open-source platform for running multiple agents together inside team channels already open in Slack, Discord, or GitHub.
Updated Aug 27
Reportedly finished Bel pre-training; actively developing Astra and GPT-6 successors
OpenAI has reportedly finished pre-training a foundation model, Bel, with more than 10 trillion parameters. If accurate, it's the company's first full-scale frontier pretraining since GPT-4o shipped in May 2024.
Updated Aug 26
Launched ChatGPT for Teens; defendant in the Raine suit
Ask ChatGPT the wrong question now, and if the system guesses you are under 18, it answers like a careful teacher instead of a confidant. On August 18, OpenAI began rolling out ChatGPT for Teens, a stricter version for ages 13 to 17 that it can switch on without being asked.
Updated Aug 18
Tenant under a 20-year lease
Nvidia sells OpenAI its chips. It has also agreed to guarantee up to $105 billion of OpenAI's rent and power bills for a huge new data center in rural Ohio, according to a securities filing made public August 17.
Maker of ChatGPT, the consumer AI market leader
A billion people now open Google's Gemini assistant each month. Google said Tuesday the app passed that mark, its 14th product to reach a billion users and, by its own count, the fastest to get there.
Updated Aug 12
Committed to 6 gigawatts of AMD GPUs
AMD began shipping Helios racks in Q3 2026, with OpenAI, Meta, Microsoft, Oracle, and Anthropic all committed as launch customers. A deal with Anthropic on July 22 added 2 gigawatts of MI450 GPU capacity and an AMD equity investment of up to $5 billion, pushing total committed deployments to 14 gigawatts. Each Helios rack lists at $5.25 million, roughly 40% above Nvidia's competing second-generation Rubin rack.
Updated Aug 5
Dominant publisher in the data set
The companies claiming to reinvent science are mostly absent from the scientific record. A study covered by Science on July 27, 2026, found that more than half of the world's AI unicorns have never led a single peer-reviewed paper or preprint.
Updated Jul 29
Entering hardware for the first time
OpenAI has never sold a physical product. On July 15, it started, with a $230 desk gadget for programmers.
Updated Jul 15
Competing on model quality and product velocity as Google expands Gemini defaults.
When Google launched Gemini 3 Flash in December 2025, it bet that a fast, cheap model could become the default brain of Search, the Gemini app, and developer tooling. That bet compounded. Google shipped Gemini 3.1 Pro in February 2026 and Gemini 3.5 Flash in May; the latter reached general availability on July 14 and is now the default across every major Google surface.
Released GPT-5.5-Cyber to vetted security teams in May 2026; scores 58.6% on SWE-Bench Pro
Anthropic restored Claude Fable 5 globally on July 1, 19 days after Commerce Secretary Howard Lutnick ordered its suspension. The company built a new safety classifier, developed with Amazon, Microsoft, and Google, that blocks the cited jailbreak in more than 99 percent of cases. Lutnick confirmed the government spent two weeks working with Anthropic to ensure 'alignment with the U.S. Government.'
Updated Jul 6
Operator of ChatGPT and co-author of the Agentic Commerce Protocol
Shopping inside AI chatbots moved from experiment to infrastructure in the first half of 2026, when Newegg launched full in-chat checkout on June 26. By then, two rival open standards were already competing for control of how AI-native checkout works.
Updated Jun 27
Cooperated with government request to stage GPT-5.6 launch to ~20 vetted companies; said customer-by-customer government approval should not be the long-term norm
Trump signed a voluntary framework on June 2 asking frontier labs to share their newest models with federal agencies for up to 30 days before release. Within ten days, the government had gone further. A June 12 export control directive forced Anthropic to kill worldwide access to Fable 5 and Mythos 5, its two newest models, after the government said it had found a jailbreak in the more capable one.
Updated Jun 26
Designed its first chip
Every time you ask ChatGPT a question, OpenAI rents the chips that answer it, mostly from Nvidia. On June 24, OpenAI and Broadcom unveiled Jalapeño, OpenAI's first chip of its own, built to do that work for roughly half the cost.
Updated Jun 25
Filed confidential S-1 on June 8; targeting September–November 2026 IPO with Goldman Sachs and Morgan Stanley as lead underwriters
SoftBank's talks to borrow $6 billion against its OpenAI shares stalled on June 10 after banks balked at pricing private-company collateral. SoftBank had gathered roughly $5 billion in soft commitments before the pause.
Updated Jun 11
Codex expanded to Windows in May 2026; Project Polaris signals Microsoft is reducing reliance on OpenAI models inside GitHub Copilot starting August 2026
OpenAI's Codex can now see and click inside Windows applications, a capability Anthropic's Claude has offered since October 2024. Microsoft joined them at Build 2026 in early June, launching its own Windows Agent Framework as open-source software and introducing Scout, an always-on background agent for Microsoft 365.
Updated Jun 8
Rolled out Dreaming V3 memory to paid users
For most of ChatGPT's life, you had to tell it to remember things, one fact at a time. On June 5, 2026, OpenAI changed that. Its new 'Dreaming V3' system reads across years of your past chats and writes its own running profile of you, no prompting required.
Updated Jun 5
Shipping a new default model to defend ChatGPT's lead
ChatGPT receives roughly a billion visits a month, and on May 5 those visitors began talking to a different model by default. OpenAI replaced GPT-5.3 Instant—the everyday workhorse it shipped earlier this year—with GPT-5.5 Instant, a faster system the company says produces 52.5% fewer hallucinated claims on medical, legal, and financial questions. The new model can also pull context from a user's past chats, uploaded files, and Gmail to personalize answers, and it appears in OpenAI's developer interface as 'chat-latest.'
Updated May 31
Reallocating compute and engineers from Sora to enterprise and coding products
OpenAI previewed Sora as a glimpse of cinema's AI future in February 2024. Twenty-six months later, on April 26, 2026, the company switched off the Sora consumer app for good.
Also preparing for IPO, targeting late 2026 at up to $1 trillion valuation
Anthropic offered employees up to $6 billion in liquidity through a tender offer at a $350 billion valuation — the same price as its February fundraising round. Employees mostly said no. The sale completed in early April well below its target because staff chose to hold their shares, betting that the company's planned initial public offering (IPO) later in 2026 will deliver a higher price.
Under state investigation and facing multiple wrongful death lawsuits
Florida's attorney general announced a formal investigation into OpenAI on April 9, 2026, alleging ChatGPT played a role in the April 2025 mass shooting at Florida State University that killed two and injured five. Court records show the shooter entered more than 270 prompts asking how the country would react to a campus shooting, when the student union is busiest, and how to operate his firearms.
Launched GPT-5.4-Cyber on April 14, a restricted cybersecurity model following Anthropic's Mythos announcement
Anthropic built an AI model so capable at finding software vulnerabilities that it decided not to sell it. Claude Mythos Preview, announced April 7, autonomously discovered thousands of previously unknown security flaws in every major OS and web browser — including a remote crash bug in OpenBSD undetected since 1999. Rather than offering the model commercially, Anthropic restricted access to 12 major technology companies through Project Glasswing, backed by $100 million in usage credits.
Updated May 30
Preparing for IPO while restructuring infrastructure strategy
Four months ago, Sam Altman told the world OpenAI had $1.4 trillion in data center commitments. Now the company is telling investors the real number is $600 billion — and that it would rather rent computing power than build its own facilities. The retreat, disclosed to investors in February 2026 and detailed publicly on March 22, marks the sharpest pivot in the short history of the artificial intelligence spending boom.
Launched Codex Security as expansion of Codex coding platform into application security
For decades, finding security flaws in software has required expensive human experts or pattern-matching tools that miss complex bugs. In five months, all three frontier artificial intelligence labs (OpenAI, Anthropic, and Google) released autonomous agents that read code like a human researcher, discover vulnerabilities traditional scanners miss, and generate patches. On March 6, 2026, OpenAI launched Codex Security in research preview, an agent that scanned 1.2 million code commits in its first month of beta testing and discovered 14 previously unknown vulnerabilities serious enough to receive formal identifiers in OpenSSH, Chromium, and PHP.
Amended Pentagon contract holder; managing consumer backlash
For decades, the United States military chose its weapons contractors and they complied, but artificial intelligence changed that. On March 3, OpenAI and the Pentagon amended a freshly signed AI contract to ban domestic surveillance of Americans—a concession the Pentagon had refused Anthropic, triggering the company's blacklist from all federal agencies.
Published the threat intelligence report; banned involved accounts
A Chinese law enforcement official used ChatGPT the way most people use a private notebook: to draft, revise, and polish status reports about their work. The problem: the work was a covert campaign to silence critics of the Chinese Communist Party living overseas. OpenAI's threat intelligence team read the reports, pieced together a transnational repression operation involving hundreds of operators, thousands of fake social media accounts, forged American court documents, and impersonation of United States immigration officials; it then published the findings.
Updated May 29
Ended Microsoft Azure exclusivity; targeting Q4 2026 IPO at up to $1 trillion; won Musk lawsuit; annualized revenue at $25B run rate
After the $122 billion round closed in April, OpenAI ended its Azure cloud exclusivity and capped Microsoft's revenue share at $38 billion through 2030. On May 18, a jury dismissed Elon Musk's suit seeking $134 billion and the reversal of OpenAI's for-profit structure.
GPT-5.2 Pro central to multiple Erdős problem solutions
For the first time, AI systems are independently solving mathematical problems that stumped human researchers for decades. Since Christmas 2025, 15 legendary Erdős problems have moved from 'open' to 'solved'—11 by AI, including the first on January 6, 2026, from OpenAI's GPT-5.2 Pro and Harmonic's Aristotle theorem prover.
Filed confidential S-1 with SEC on May 22, 2026; targeting Q4 2026 IPO at up to $1 trillion; reported negative 122% operating margin in Q1 2026
On May 28, Anthropic closed a $65 billion Series H at a $965 billion valuation, passing OpenAI to become the world's most valuable private AI company. The round disclosed annualized revenue at $47 billion, up from $44 billion three weeks earlier. Anthropic is targeting an IPO this autumn.
Primary competitor in reasoning model development
OpenAI launched the first commercial reasoning model in September 2024. Seventeen months later, Google's upgraded Gemini 3 Deep Think pulled ahead on the benchmarks that matter most for science.
Released GPT-5.5 (April 23); annualized revenue ~$24-25B, trailing Anthropic's reported $30B ARR; committed $250B in Azure contracts to Microsoft; $2B monthly burn; pivoting fully from owned infrastructure to rented cloud compute
The four largest cloud providers—Microsoft, Meta, Alphabet, and Amazon—are tracking toward over $720 billion in combined AI infrastructure spending for 2026, up sharply from $410 billion in 2025. All four reported first-quarter results on April 29, 2026. Microsoft delivered the clearest signal: $77.7 billion in revenue (up 18% year-over-year), Azure cloud growth of 40% (above its 37% guidance), and earnings per share of $4.13 versus analyst estimates of $3.67.
Updated May 26
Produced the disputed October 2025 claim and the verified May 2026 proof
For 80 years, mathematicians believed square grids were the best way to pack points at distance exactly 1 apart on a plane. On May 20, an internal OpenAI reasoning model produced a counterexample, using algebraic number theory to beat the grid by a small polynomial factor. Nine outside mathematicians verified the proof.
Updated May 24
Issuer of the confidential IPO filing
OpenAI filed a confidential draft registration with the Securities and Exchange Commission (SEC) on Friday, May 22, 2026. The filing is the first formal step toward a public listing that could value the ChatGPT maker above $1 trillion.
Updated May 22
Launching its first ad business
OpenAI ran ChatGPT as a paid and API-only product for three years. On May 21, the company opened a self-serve ad platform that lets any business buy placements inside the chatbot, with no minimum spend.
Updated May 21
Lobbying for federal preemption of state laws
The DOJ's AI Litigation Task Force began operations January 10, 2026 with one mission: kill state AI laws in federal court. Attorney General Pam Bondi's team, consulting with AI czar David Sacks, will challenge comprehensive AI regulations from California, Texas, and Colorado that President Trump's December executive order called unconstitutional burdens on interstate commerce.
Updated May 20
Main rival on agentic AI
Sundar Pichai opened Google I/O 2026 with a Gemini model that can run a user's desktop. The keynote unveiled Android XR smart glasses too. Aluminium OS, a laptop platform merging Android and ChromeOS, will ship as 'Googlebooks' from Acer, ASUS, Dell, HP, and Lenovo.
Updated May 19
Lost enterprise lead to Anthropic; share fell to 32.3 percent in April 2026 — a second straight monthly decline — on Ramp's data
Ramp's May 2026 AI Index shows 50.6 percent of US businesses now pay for AI tools, up from 50.4 percent in March. Anthropic holds 34.4 percent of those paying customers against OpenAI's 32.3 percent — the second consecutive month Anthropic has led.
Updated May 15
Secured Disney as Sora’s first major content partner and a $1B strategic investor.
Mickey Mouse just shook hands with the algorithm Hollywood spent two years trying to tame. Disney is investing $1 billion in OpenAI, letting Sora and ChatGPT Images legally generate short videos and images of more than 200 Disney, Pixar, Marvel, and Star Wars characters — not the actors who play them.
Updated May 11
No stories match your search
Try a different keyword
How would you like to describe your experience with the app today?
