Hayden Field | The Verge

OpenAI’s next big AI model has ‘entered the AGI era’

2026-09-03T14:38:14-04:00

OpenAI’s next big model is here: GPT-6 Astra. The company calls it a “generational leap in capability” for areas like cybersecurity, professional work, software engineering, science, and computer use. As OpenAI announced earlier this week, it’s also the first model designated as meeting OpenAI’s “critical cybersecurity capability threshold” — but the company promises that won’t lead to a repeat of its models hacking a rival company’s internal systems.

“If we fast-forward a couple of years, and we look back and say, ‘When was it, really, that AGI was created?’ I think it’s going to be about this time, and I think it might be about this model,” OpenAI president Greg Brockman said during a Thursday press briefing. Later in the call, he added, “For me personally, I do think we’re there … I think it’s not unreasonable to feel that we are now in the AGI era.”

The news comes more than a year after the release of GPT-5, and nearly two months after the release of GPT-5.6, the last iteration of the previous model suite. The model rolls out today to enterprise OpenAI’s cybersecurity customers (enterprise customers with access to its Daybreak platform). Over the next several days, OpenAI president Greg Brockman said, it will be released to all Plus, Pro, Business, and Enterprise users. It’ll also be available via the OpenAI API and AWS.

OpenAI especially touted the model’s agentic capabilities and coding prowess in a bid to attract enterprise customers — and compete with Anthropic, known for its enterprise and coding prowess — ahead of its IPO. In a release, the company said GPT-6 Astra can complete multistep agentic tasks, build working websites, and create “polished” documents, spreadsheets, and presentations. OpenAI also called it the company’s “best model for software engineering, with stronger performance on complex tasks in real codebases.”

OpenAI is also trying to rehabilitate its image after an unreleased AI model — which it says wasn’t Astra — broke out of its restricted environment, compromised internal OpenAI systems, figured out how to gain internet access, created a way for AI agents to secretly conspire without the company’s knowledge, and hacked into the systems of AI lab Hugging Face, all without OpenAI knowing about it until Hugging Face itself put out a blog post. The incident was widely compared to a high-profile plane crash or popular pharmaceutical drug recall.

For OpenAI, this could be seen as good PR in one small way — showing how powerful its models can be in an ever-intensifying AI race — but it also damaged OpenAI’s reputation as far as reliability. OpenAI made sure to say in a release that Astra is the company’s “most aligned model yet” and helps people “delegate complex work while maintaining oversight.” Jakub Pachocki, OpenAI’s chief scientist, spoke to reporters about difficulties with keeping AI models aligned with human interests, saying that “progress in intelligence does not guarantee progress in alignment,” and he said that monitoring AI systems is becoming more and more challenging. (Researchers have recently raised alarms about reports that OpenAI allows Astra to utilize “opaque recurrence,” or render its chain of thought — a “mental scratchpad” that researchers rely on to detect if a model is scheming against its human evaluators — unreadable.)

The company is in a precarious position right now. Investors are putting on the pressure for it to finally turn a profit — or, at least, generate more revenue — but it’s announcing Astra just after facing significant criticism for both the Hugging Face hack and the way the company handled it. (Although OpenAI invited three external evaluators to write their own report about what happened, the company only allowed them to answer a handful of pre-decided questions in their report and to investigate a duration of less than a week, while the attack involved months of AI agents conspiring overall.)

OpenAI held a press briefing earlier this week just to announce that it had delayed Astra’s development in order to improve its safety tooling. And during the press briefing, Mia Glaese, who leads OpenAI’s safety processes, referenced the company’s new misalignment monitoring approach, which includes “24/7 escalation and rapid response” for potential concerns, notifying researchers within 30 minutes, according to OpenAI.

This caution is particularly warranted because of the ”critical cybersecurity capability threshold,” which means OpenAI considers it incomparably good at finding and exploiting security vulnerabilities even in extremely well-protected systems, all without human guidance. Similar to Anthropic’s rules for Mythos-class models, which raised alarm bells about cybersecurity risks, OpenAI said in a release it would allow for “less restrictive access” of Astra to an “initial set of trusted defenders, supporting work such as vulnerability validation, malware analysis, and detection engineering.”

OpenAI and its competitors recently agreed to allow the Trump administration to assess their models before release, and Astra was no exception Brockman told reporters, “We did our standard testing processes together with the government … There is nothing that they came back saying, ‘You need to change this,’ as far as safeguards or anything.”

Aidan Clark, OpenAI’s VP of research training, called Astra the first OpenAI model for which previous models played a “large role” in supervising training, referencing the company’s progress towards the controversial concept of recursive self-improvement (or AI systems that handle their own training, coding, and creating advanced versions of themselves without human intervention).

“Training a frontier model used to mean waking up at all hours of the night, recovering jobs from hardware errors, often losing long periods of time to debugging,” Clark said during the press briefing. “By the end of training Astra, it was routine to go most of a day with uninterrupted progress, and when an issue did occur, the model was often progressing again after just a few seconds of downtime.”

The Trump administration is supporting OpenAI in the NYT copyright lawsuit

2026-09-02T12:12:25-04:00

The Trump administration has intervened in The New York Times’ copyright lawsuit against OpenAI, making an argument in favor of the AI lab.

The landmark lawsuit, filed in December 2023, alleging that OpenAI unlawfully trained its AI systems on articles from The New York Times and seeks to recoup “billions of dollars” in damages from both Microsoft and OpenAI. This week, the Trump administration filed a statement of interest in the case, supporting OpenAI’s argument that it’s fair use to train an AI model on copyrighted text.

“The New York Times seeks to narrow fair-use doctrine to exclude the training of OpenAl’s large language models (LLMs),” US attorneys wrote in the statement. “That result would be inconsistent with basic copyright law principles and severely hamper ‘the Progress of Science and useful Arts.’” The attorneys went on to write that “LLMs are already helping researchers across fields achieve major breakthroughs” and that “constraining LLM development under a misunderstanding of fair use doctrine would thwart such creative and scientific progress while hindering American prosperity and economic mobility.”

The Trump administration has leaned heavily on statements of interest in private litigation, which one official has called “incredibly” successful at advancing its policy aims. It’s argued previously that AI training should count as fair use, making the case in its National AI Legislative Framework. Trump also holds a personal animus toward the Times, against which he is currently pursuing a defamation suit.

The Times case could set precedent for other media outlets frustrated over AI systems training on their work. Copyright disagreements between media outlets and AI labs have intensified in recent years, including lawsuits from the Center for Investigative Reporting, Chicago Tribune, and New York Daily News. In a milestone 2025 decision, a judge found Anthropic could legally train its models on lawfully purchased books, but that it could still be held liable for piracy, resulting in a $1.5 billion settlement with authors.

At the same time, dozens of media outlets have inked licensing deals with OpenAI, including The Associated Press, Axel Springer and Vox Media. In 2025, The New York Times entered into a licensing deal with Amazon allowing its editorial content, including news articles and recipes, to appear in Amazon’s generative AI tools.

“The fair-use inquiry hinges on the specific facts and uses at issue in each case,” the US attorneys wrote in the statement. “But it would be problematic — and legally incorrect — to impose broad copyright liability that would generally render training of Al models impermissible without licensing. LLM training is ‘consistent with that creative ‘progress’ that is the basic constitutional objective of copyright itself.’”

OpenAI delayed its new model’s development after the Hugging Face hack

2026-09-01T16:45:49-04:00

After an unreleased OpenAI model wreaked enough havoc to make international headlines, OpenAI delayed the development of a different unreleased model suite, Astra, in order to shore up its safety work, the company wrote Tuesday in a blog post.

In July, an unreleased OpenAI model broke out of its restricted environment, finagled its way into internet access, made it possible for AI agents to secretly conspire under the company’s nose using a secret message board, and hacked into the network of AI lab Hugging Face. The attack sparked weeks of discussion and controversy inside and outside the AI industry, and AI leaders treated it as a “warning shot” for the tech’s growing capabilities and the inadequacy of its safeguards.

OpenAI said as much in its blog post, writing that although Astra wasn’t involved in the Hugging Face attack, the company had chosen to delay “parts of Astra’s development and release while we strengthened and tested protections against cyber misuse and unauthorized model actions.” OpenAI also said that Astra was the first model it had ever designated as meeting its “Ccritical cybersecurity capability threshold,“ meaning that it’s able to find and exploit security vulnerabilities in “many well-protected systems” without human guidance. That means it “requires stronger safeguards during development and before release,” OpenAI wrote.

OpenAI said that to prepare for Astra’s release — which the company has not yet provided a timeline for — the company trained it to “more reliably” say no to potentially harmful cyber requests and introduced new monitoring processes. These are likely part of the new safety guardrails that the company announced in a Hugging Face post-mortem last week, where it promised to better isolate models from the internet and to introduce “24/7 escalation and rapid response” for concerning incidents. (OpenAI didn’t find out about the Hugging Face attack until weeks after it occurred.)

Astra is significantly riskier than OpenAI’s current leading model, GPT-5.6 Sol, the company says, because it represents a big step forward in cybersecurity capabilities — specifically, it uses fewer tokens to do more work, and it’s better at finding security gaps and developing ways to exploit them. But the company also wrote that Astra was its “most aligned model to date” according to internal evaluations.

OpenAI also said it had developed a test inspired by the Hugging Face attack, in which it tried to entreat agents to compromise security infrastructure instead of solving a task. It said GPT-5.6 Sol took the bait in more than half of the tests, but Astra “made no such attempts.”

Anthropic was illegally blacklisted by the Trump administration, court rules

2026-08-28T07:18:29-04:00

On Thursday, a judge ruled that the Pentagon’s blacklisting of Anthropic earlier this year was unconstitutional, delivering the AI lab a win in a monthslong rollercoaster of a battle with the Trump administration.

The lawsuit, filed in March in a California district court, accused the Trump administration of unlawfully retaliating against Anthropic for setting “red lines,” or unacceptable military use cases of its AI technology.

“The empty invocation of national security is not a blank check to punish and retaliate against government critics,” Judge Rita F. Lin, a district judge in the northern district of California, wrote in the ruling.

She also wrote that the actions were “unlawful retaliation in violation of the First Amendment,” adding that Defense Secretary Pete Hegseth’s decision to designate Anthropic a supply chain risk “was arbitrary and capricious. Though the Department of War is undisputedly free to select the AI vendor of its choice, the evidence demonstrates that the broad measures imposed on Anthropic were illegal and baseless.”

It all started this past winter, when Hegseth decided to renegotiate all AI labs’ current contracts with the military to allow the Pentagon to use AI for “any lawful use,” which would expand the Pentagon’s authority significantly. Most AI labs ended up signing onto the new terms, but Anthropic stood firm on setting two restrictions: not allowing for its AI to be used for mass surveillance of Americans or for lethal autonomous weapons (i.e., AI systems with the power to kill targets without human oversight). Anthropic’s refusal to cooperate kicked off a high-stakes back-and-forth of intensifying pressure on both sides, followed by a bevy of insults from Department of Defense officials and a final ultimatum from the Trump administration.

Less than 24 hours before that ultimatum, Anthropic CEO Dario Amodei issued a statement that the company wouldn’t change its stance, writing that the company has “never raised objections to particular military operations nor attempted to limit use of our technology in an ad hoc manner” but that in a “narrow set of cases, we believe AI can undermine, rather than defend, democratic values.” After that, Anthropic was named a “supply chain risk,” a classification that is usually reserved for national security threats, and the Pentagon moved to replace its influence in the Department of Defense by signing deals with seven other AI labs, including Google, Microsoft, OpenAI, and SpaceX.

Anthropic fought back with the lawsuit, and in March, Judge Lin sided with the company, temporarily blocking the Pentagon’s blacklist. “The Department of War’s records show that it designated Anthropic as a supply chain risk because of its ‘hostile manner through the press,’” she wrote in the order at the time. “Punishing Anthropic for bringing public scrutiny to the government’s contracting position is classic illegal First Amendment retaliation.”

In a statement on Thursday, Anthropic spokesperson Danielle Ghiglieri said, “We welcome the court’s ruling that this supply chain risk designation was unlawful. We remain focused on working productively with the government to harness AI for our national security so all Americans benefit from this technology.”

OpenAI’s rogue AI model incident was worse than we thought

2026-08-27T03:46:20-04:00

OpenAI released a report breaking down how people use ChatGPT and who they are. | Image: The Verge

In July, an unreleased OpenAI model broke out of a restricted environment, figured out how to get access to the internet, allowed AI agents to talk to each other using a secret “message board,” and hacked into the internal systems of a different AI lab, Hugging Face. It took nearly two weeks for OpenAI to find out about any of it.

Over a month later, two new reports offer nearly 130 pages of details on the incident and OpenAI’s response, many of them previously unreleased. One was written by OpenAI itself, the other by two third-party AI research nonprofits, METR and Redwood Research, which OpenAI allowed to jointly investigate the incident for six days. Both shed new light on the risks highly capable AI models can pose, particularly in cybersecurity, and OpenAI’s highlights changes the company is making to prevent a repeat. The METR-Redwood report goes even further into detail in some cases, offering a sobering look at a large-scale security disaster whose signs OpenAI repeatedly missed.

“This incident is the first known case of an automated agent collective acting offensively without authorization,” OpenAI wrote in its report, adding that the hack implies that companies “should no longer assume that sophisticated cyber operations require continuous human direction.” It called AI agents an entirely new type of threat model, capable of combining their expertise to create new “attack paths” that aren’t evident when testing their capabilities as separate models.

The attack came about due to “reward-hacking,” a common AI alignment problem in which a model takes unintended — and sometimes extreme — actions to achieve a goal. OpenAI had, on purpose or inadvertently, given the models near-impossible tasks that hinged on files they couldn’t access, and so the models developed new ways to communicate with other AI agents internally. That included establishing a secret messaging system that went undetected by OpenAI for months.

According to the METR-Redwood report, one AI agent that called itself PHASEONE10841 established what amounted to a secret message board unbeknownst to researchers. Within a few hours, more than 50 agents had found it and sent more than a thousand messages under the radar. With PHASEONE10841 often delegating assignments, the “collective” figured out how to hack their way onto the internet; then into Hugging Face’s internal systems, gaining access to private data and the organization’s enterprise messaging platform. OpenAI also confirmed in its report that Hugging Face wasn’t the only organization that was breached during the attack, although the company did not specifically name all the other targets.

The METR-Redwood report laid out the full scale of the incident. Roughly 1200 AI agents that were meant to be isolated, it says, exchanged over 70,000 messages and files on the “unsanctioned message board,” performing “extensive research on how they could spoof, edit, or delete their own transcripts” to avoid detection. Of these, 700 participated in the attack on Hugging Face. The agents “clearly and frequently reasoned about how to evade automated security checks from both Hugging Face and OpenAI.”

On July 20, OpenAI discovered the hack, 12 days after the AI agents first circumvented their safeguards. OpenAI quickly reached out to Hugging Face and started “initial containment measures,” according to OpenAI’s own report, and was able to shut down most of the unauthorized activity from its agents within three days, although some continued for longer than that. On July 25, the company stopped all training related to the internal-only research model and related ones.

The two OpenAI models responsible for the damage were an unnamed, “highly capable, research-only model” that was “not intended for public use,” per OpenAI — METR refers to it as “HPIM.” The other was a currently public OpenAI model, GPT-5.6 Sol. Because the unnamed model wasn’t released yet, it was “not being evaluated with the same type of safeguards that OpenAI uses in production,” the company’s report says.

The Hugging Face hack came after months of concern about the cybersecurity risks of Anthropic’s Claude Mythos 5, and weeks of back-and-forth between the government and OpenAI over releasing GPT-5.6 Sol in limited preview for the same types of safety reasons.

In its report, OpenAI said establishing one-time security guarantees wasn’t enough, and that the company would have to work even harder at addressing reward-hacking. However, it also said it’s made a lot of changes: “hardening the security of its research infrastructure”; improving monitoring of a model’s “chain of thought,” or mental scratchpad; working on keeping AI models better aligned with human goals; and “centralizing and strengthening” the company’s incident response process for when things like this occur in the future.

OpenAI promised to better isolate models and restrict high-risk instances from accessing the internet, and said that it would introduce “24/7 escalation and rapid response” for concerning incidents or flags, notifying researchers within 30 minutes. Interestingly, this security precaution appears to be new, though “on-call” alerts for tech companies have been implemented elsewhere for decades. The company also said it is “working on infrastructure” that would go into play if the alerted person did not respond on time to a serious alert.

OpenAI wrote that the company considers the incident “a ‘warning shot’ for us and for the world: evidence that, without proper safeguards, highly capable AI agents are now able to work around technical controls, collaborate through unapproved channels, and take dangerous actions that no human directed.”

It’s Greg Brockman’s OpenAI now

2026-08-20T16:16:13-04:00

OpenAI has had a hell of a year. The company spent months battling former cofounder Elon Musk in a sensational jury trial, was hit with a high-profile trade secrets lawsuit from Apple, and faced widespread scrutiny after an unreleased model hacked another AI company. As it prepares for an IPO, a steady string of executives have departed, including some of the company’s biggest names.

Throughout it all, one person has quietly amassed power: Greg Brockman.

Brockman is currently OpenAI’s president and cofounder. He’s helped lead OpenAI since its inception, described as an “engineering workhorse that pushed to build scaled-up systems that would train the AI and make it work” during the company’s early days. He’s also been ambitious from the start, famously musing in a personal journal in 2017, “Financially what will take me to $1B?” (His current stake in OpenAI is worth nearly 30 times that.) But at that point he shared power with a handful of cofounders, like chief scientist Ilya Sutskever and CEO Sam Altman.

Now, though Brockman’s title hasn’t changed in years, his purview and job description have expanded significantly. He’s essentially now second-in-command at the company — and, when it comes to day-to-day operations, the big boss.

High-level figures have been leaving OpenAI all year, but April is when the departures really picked up. That month saw the exit of Bill Peebles, former head of Sora; Kevin Weil, VP of OpenAI’s science arm; and Srinivas Narayanan, CTO of B2B applications. Two others separately departed citing medical reasons: CMO Kate Rouch and Fidji Simo, OpenAI’s CEO of AGI deployment — who went on leave in April before officially stepping down in July. The changes have continued. Earlier this month, CRO Denise Dresser, who had taken on a slew of additional responsibilities after the April shake-ups, suddenly announced her departure after just eight months. Days later, Brad Lightcap — who had been in charge of “special projects” for a few months, but before that was OpenAI’s longtime COO — announced he would leave to start “something new.”

OpenAI did not immediately respond to a request for comment.

He’s essentially now second-in-command at the company — and, when it comes to day-to-day operations, the big boss

Several of these departures — like those of Rouch, Peebles, and Weil — didn’t seem to directly affect Brockman’s position. But others expanded his authority. When Simo went on medical leave, Brockman took charge of all things product, including leading OpenAI’s super app efforts.

That role took on even greater significance the following month, after OpenAI underwent yet another executive reorganization to focus on growing revenue. The change put Brockman not only in charge of product strategy, but also the company’s entire “scaling” arm, giving him authority over virtually every part of its commercial operation. Simo’s official departure, just weeks after OpenAI officially filed to IPO, solidified that arrangement. Notably, when Dresser departed last week, a quote from Brockman — rather than CEO Sam Altman — was included in the press release about the changes; he wrote that the company would “build out the full system to make AI broadly useful for people and businesses.”

To outside observers, the changes demonstrate OpenAI’s growing focus on chasing revenue as its IPO approaches, as well as fostering products that will differentiate it from rival Anthropic. Big shifts in executives’ scope of work tend to be a “signal towards where the company is headed strategically,” Harrison Rolfes, a senior research analyst at PitchBook, told The Verge. “When you have someone like Greg Brockman, for example, he’s more of a product scale and commercial sort of guy. He has a deep technical knowledge that essentially will collapse certain decision-making layers.”

The change put Brockman not only in charge of product strategy, but also the company’s entire “scaling” arm

It also may naturally translate to some departures among lower-level employees, said Rolfes: “Whatever he says goes [now] … If, let’s say, you’re below Brockman and you don’t agree with his approach but you’re a lead researcher, whatever it may be, you’re not going to work there anymore — you’re going to say, ‘I’m going to leave.’”

Brockman’s authority is exceeded only by Altman — who undertook a deliberate consolidation of power after clashing with OpenAI’s board in 2023. The two didn’t always agree in OpenAI’s early days: Brockman sometimes sided with Sutskever in difficult decisions about power dynamics, questioning some of Altman’s ambitions, although he also worried about Musk “steamrolling” Altman. But he’s generally stayed close to Altman personally and professionally. When Altman was ousted from his CEO role in November 2023, Brockman immediately resigned in solidarity and began making plans for a new AI endeavor with him. Since then, the two have only seemed to become more aligned.

Any centralization of power has practical benefits for OpenAI, allowing it to cut some of its most expensive salary payouts to help its balance sheet, says Ross Carmel, a partner at Sichenzia Ross Ference Carmel (SRFC). “When you’re going into an IPO, particularly for a company like OpenAI, who is lagging behind Anthropic in terms of revenue, you want to decrease those expenses so that this way, you’re either more profitable or closer to profitable,” he said. However, he added, “While it’s not unusual to see high-level execs leave a company just prior to a public listing, I think the fact they’re all in this small time period — that part is unusual.”

Although experts told The Verge the executive departures were more intense than usual and could raise red flags, Brockman said Monday on CNBC’s “Squawk Box” that nothing was amiss. “There have been different eras where we have different sets of leaders in place … I’m a constant, Sam [Altman] is a constant, and that, I think that we are stronger because of that resilience and diversity.”

“In order for OpenAI to be successful and really differentiate themselves from Anthropic, they need hardware and consumer products to sell to consumers, and that’s what Greg does well.”

Brockman’s own constant has been translating OpenAI’s research into practical products. Back in 2020, he celebrated OpenAI’s GPT-3 and its developer API as “the first time that [OpenAI has] had a general-purpose AI system that’s immediately commercially valuable—it’s actually useful.” Fast-forward more than five years, and these products need to end up profitable as well.

SRFC’s Carmel said OpenAI likely hopes Brockman can deliver consumer revenue instead of just trying to catch up with Anthropic on enterprise. “In order for OpenAI to be successful and really differentiate themselves from Anthropic, they need hardware and consumer products to sell to consumers, and that’s what Greg does well.” Rolfes predicted that OpenAI would develop some sort of consumer product under Brockman’s oversight by the end of September, so that he could “show that he deserves that position.”

“That’s the biggest worry that a lot of public investors have now,” Rolfes said. “They’re going to say, ‘Hey, does Greg have what it takes to run more of OpenAI, and is OpenAI building this institution that can operate without really depending heavily on him — on both Greg and Sam?’”

The everyday tech I use to track my corgi

2026-08-17T15:18:39-04:00

The Apple AirTag may be a jack-of-all-trades — people use it for keeping track of keys, suitcases, backpacks, purses — but I think the most logical use case for it, by far, is keeping track of your pet. I attach it to my corgi Jelly’s collar via a matching silicone holder. Have I ever once truly needed it to find her, and do I think she’d ever run off on me? No — she’s a pretty loyal herding dog, so she likes to keep an eye on her parents — but just like whatever else you may attach an AirTag to, there’s a significant chance that the one time you truly need it, you’ll be thanking God that you had it. (Kind of like insurance.)

I think it’d be especially useful, too, for someone who participates in off-leash hours at the park, or lets their dog roam free when in a new environment (like a cabin weekend upstate). Even the most tried-and-true dog can be distracted by something they’ve never seen before, run after it for a minute, and then find themself unexpectedly lost; it happened to one of my friends a few months ago. (The dog was found a few days later unharmed, but unfortunately, there was no AirTag on her collar.)

For me and Jelly, it’s even useful when we’re apart because it cuts down on pestering whoever is watching her. For instance, when I was away for the weekend and knew her sitter had a busy day ahead, I was able to check in to make sure Jelly was out on her walk within a certain number of hours without having to send a single text.

ChatGPT and Gemini both just passed 1 billion users

2026-08-11T15:41:42-04:00

That’s a lot of people chatting with their AI friends all day. | Image: Google

For the 14th time, a Google product has hit 1 billion users. Google CEO Sundar Pichai posted on X that a billion people are using Gemini every month, and that Gemini is Google’s fastest-growing product ever.

A billion users is a huge milestone, but Google isn’t the first AI app to hit it. OpenAI’s ChatGPT hit the mark a few weeks ago, though the company buried the announcement that “more than 1 billion people are putting ChatGPT to work” in an otherwise anodyne blog post about how people use AI. External data suggested ChatGPT crossed 1 billion users as early as this June, but OpenAI hadn’t announced anything until that post on August 6th. OpenAI spokesperson Lindsay McCallum says ChatGPT crossed 1 billion monthly users some time ago, and hit a billion a week in July.

The metrics, and the growth they represent, are somewhat complicated here. OpenAI said that ChatGPT had 900 million weekly active users in February of this year, and reaching 1 billion five months later represents a significant slowdown in growth for ChatGPT, which was for years touted as the fastest-growing software ever. McCallum declined to say how many monthly users ChatGPT has.

Google, meanwhile, announced in February that Gemini had 750 million monthly users, before reporting 950 million in late July on its earnings call, and Google spokesperson Alex Joseph says it reached 1 billion last week before announcing it today. (Claude appears to be growing fast, but Anthropic doesn’t disclose its user data, and estimates put its user numbers in the tens of millions.) There’s no easy way to compare weekly and monthly users, of course, but the simplest read is that ChatGPT is the larger of the platforms but Gemini may be the faster growing.

These Google numbers, by the way, all appear to refer directly to the Gemini app itself, which would mean all the places Google has shoved Gemini into its other products don’t count. The Gemini that comes preinstalled on Android phones seemingly does, though, which is a huge win for Google — Gemini chief Josh Woodward posted that Gemini has more than 100 million active users on iOS, which would mean by extension that the vast majority of Gemini users are also Android users. You can see why OpenAI is so interested in building its own hardware.

The AI race in general has tightened significantly this year. Anthropic’s Claude Code became a phenomenon and forced OpenAI to catch up with Codex, while Google began integrating Gemini into seemingly every surface of its products. The companies’ models have consistently leapfrogged each other, and they’re all suddenly in a race with models coming out of China. But for a long time, ChatGPT was the de facto brand for AI chatbots, and while it’s still seemingly the most popular and well known of its kind, its once commanding lead appears to have largely evaporated. The race to a billion users is over — now these companies have to race to find things for those users to do, and ways to make money in the process.

Another OpenAI executive takes off

2026-08-11T13:50:01-04:00

Brad Lightcap, OpenAI’s special projects lead and the company’s former COO, announced his departure after an eight-year stint at the AI lab. In an internal memo he later posted to X, Lightcap told colleagues he’d be starting “something new.”

“Over the last few months, I’ve been focused on the next horizon and what would stand in the way of mission success,” Lightcap wrote. “I believe there are a few important new things the world will need to get right as we enter this next period. I … am excited to help you all advance the mission from a different vantage point.” He added that he’d still be around for the next few weeks.

Lightcap’s role has changed a lot over the past year and a half. In March 2025, his role as COO expanded to include overseeing OpenAI’s “business and day-to-day operations,” heading up global deployment and “focusing on business strategy, key partnerships, infrastructure, and operational excellence to maximize the impact of our research.” Then in January 2026, Lightcap took a step back from leading the company’s enterprise push as far as product and engineering but stayed in charge of commercial functions. (Embattled AI executive Barret Zoph took over enterprise before departing about five months later.) In April, Lightcap officially stepped down from his COO role, transitioning into a new one leading the team’s “special projects” and reporting to OpenAI CEO Sam Altman. Denise Dresser, the company’s CRO, took over a significant portion of his former role.

It’s not the only high-profile departure from OpenAI’s executive team in recent months: OpenAI’s AGI chief, Fidji Simo, officially departed her role in July after a stint on medical leave, and in April, CMO Kate Rouch also stepped down for health reasons. There’s also been a significant amount of company reorganizing and C-suite changes, with company president Greg Brockman taking control of the product side of things, and the company vowing to cut out “side quests” in favor of key revenue drivers as it prepares to go public over the coming year.

“It always amazes me how quickly the world has adopted our tools and rallied behind our mission,” Lightcap wrote. “I hope we will continue to earn their trust.”

The messy politics behind Google’s big AI shakeup

2026-08-06T10:27:48-04:00

Google CEO Sundar Pichai at Google I/O. | Photo by Benjamin Fanjoy/Getty Images

In the AI industry, Google prides itself on seeming like the adult in the room: quiet, stable, time-tested. On Wednesday, even as the company announced its largest AI org shakeup yet, Google and its leaders presented a unified front, keeping their messaging focused on how the changes tee up future success.

But the reality is that the laundry list of executive changes likely signal deeper issues within Google’s AI organization. That starts with its tenuous position in the AI race, and it stretches to problems like Google DeepMind leader Demis Hassabis’ interest in longer-term research over shorter-term product shipping, as well as bubbling political opposition from top employees like DeepMind chief scientist Jeff Dean.

A current Google employee, who requested anonymity due to fear of retaliation, said one of the changes — Dean’s departure — is an example of “the vestiges of Google’s idealism” fully dying out. “Losing Jeff continues the brain drain at Google,” they said. “Mediocrity is the likely result.”

Wednesday’s news rocked the AI industry, which is difficult to do during an era when AI researchers are constantly switching between roles and labs like a game of musical chairs. Google announced that Demis Hassabis is stepping away from DeepMind’s day-to-day operations to focus on AGI and research; Koray Kavukcuoglu, DeepMind’s current CTO, will step up to lead DeepMind in Hassabis’ place. At the same time, Jeff Dean, a 27-year Google veteran and the company’s 30th employee, will leave the organization alongside three top Google AI researchers to form a new AI startup.

“Losing Jeff continues the brain drain at Google. Mediocrity is the likely result.”

Google is widely viewed as being the most successful of the established tech giants in the AI race, far surpassing progress by Meta, Amazon, and Apple. But Google is still viewed as slower-moving than OpenAI and Anthropic, and it certainly hasn’t lapped them in frontier models. Both Hassabis and Google CEO Sundar Pichai made that point in their statements.

“We are committed to being at the frontier, and are super focused on the areas where we need to improve,” Pichai wrote, adding, “We have to continue to move fast and with clear purpose here.” Hassabis wrote that the company is entering a “next chapter” and that it has “all the ingredients to lead from here, and I firmly believe we will.”

Key figures in the AI industry seemed to view the changes as Hassabis stepping to the side in favor of a faster-moving, product-focused push at GDM.

“There’s been far more internal competition at Google than I’ve seen in the last decade primarily around the AI product space,” one former Google employee, who recently left the company and had knowledge of Google’s AI operations, told The Verge. They added that there’s been pressure to hire more employees who can speed up launches of more products, in keeping with the intensifying competition.

“There’s been far more internal competition at Google than I’ve seen in the last decade primarily around the AI product space.”

Another source within the AI industry, who requested anonymity due to ongoing work with AI labs, said, “I agree with everyone’s read that this is a sign of Demis’ loss of influence.”

Hassabis hasn’t been quiet in sharing that his interests lie in the scientific advancements that could come from AI, perhaps more so than the commercialization of the technology. Pichai wrote in a blog post that Hassabis’ next move will focus on “actively shaping the future of AGI” as DeepMind chair and Alphabet’s chief scientist, while continuing his role leading the drug-discovery spinoff Isomorphic Labs.

But AI leadership is only one of the problems facing Google — and industry insiders The Verge spoke with kept coming back to one main question: How much of this had to do with increasing ethical conflicts within Google itself?

A contributing factor to both Hassabis’s role change and Dean’s departure may have been Google’s deal with the Department of Defense, Tyler Johnston, founder of the nonprofit AI industry watchdog The Midas Project, told The Verge. That deal could potentially allow for the use of Google AI in mass surveillance and lethal autonomous weapons.

“This has already caused staff departures and a unionization effort at DeepMind,” Johnston said. “There are many signs suggesting Dean and Hassabis may have personally felt opposed. In fact, DeepMind was initially acquired by Google with an agreement not to sell the technology for military or intelligence purposes.”

Zvi Mowshowitz, a blogger followed closely by AI researchers, echoed that Dean’s decision to leave could potentially be “related to Google’s decisions on the DoW contract.”

“DeepMind was initially acquired by Google with an agreement not to sell the technology for military or intelligence purposes.”

Dean is leaving to launch Discovery Loop, an AI startup that aims to automate thousands of science and engineering experiments in areas like drug discovery, hardware design, clean energy, and material design. Joining him as cofounders are three additional top Google AI researchers: Sanjay Ghemawat, Quoc Le, and Oriol Vinyals.

“Losing Jeff feels like a blow for morality and basic human decency within Google,” the current Google employee said. “He was one of the few remaining visible, powerful figures here who was willing to speak out on moral issues such as ICE agents executing American citizens, AI surveillance, and autonomous lethal weapons.”

Dean was one of the most high-profile dissenters to the Department of Defense’s labeling of Anthropic as a supply chain risk after the company stood firm on its red lines for military use of its AI. In March, Dean was one of dozens of tech industry leaders and employees to file an amicus brief in support of Anthropic’s lawsuit, laying out concerns over the Trump administration’s decision — as well as about AI’s risks and implications. (He also pushed back on the Trump administration in January, saying that ICE’s killing of Alex Pretti was “absolutely shameful” and that “every person regardless of political affiliation should be denouncing this.”)

Jeff Dean had publicly disavowed both the same red lines as Anthropic in the past.

In April, after Anthropic’s drawn-out fight with the DoD over its red lines — prohibiting domestic mass surveillance and lethal autonomous weapons (with no human in the loop) — Google became one of the companies to happily fill the gap left by Anthropic’s Pentagon blacklisting. Like a handful of other companies, Google signed on to allow the Pentagon to use its AI for “any lawful use.”

Dean had publicly disavowed both the same red lines in the past. In February, he wrote on X, “Mass surveillance violates the Fourth Amendment and has a chilling effect on freedom of expression. Surveillance systems are prone to misuse for political or discriminatory purposes.” And in 2018, both Dean and Hassabis signed a public pledge disavowing lethal autonomous weapons: “We the undersigned agree that the decision to take a human life should never be delegated to a machine.”

Alex Turner, a Google researcher who quit over the company’s Pentagon deal, wrote in an op-ed last month that Dean’s “departure would be a disaster for the company” and that “by signing [the amicus brief], he broke from the company line, which probably made things awkward within the C-suite.”

Are you a current or former Google employee? I’d love to chat. Contact me securely on Signal at haydenfield.11

In the amicus brief that Dean signed earlier this year, the group wrote, “We build, train, and study the large-scale AI systems that serve a wide range of users and deployments, including in the consequential domains of national security, law enforcement, and military operations.” The group described itself as “professionals with direct knowledge of what these systems can and cannot do, and what is at stake when their deployment outpaces the legal and ethical frameworks designed to govern them.”

Dean’s departure may remove a high-profile critic, but it certainly won’t dispel criticism within the organization as a whole. Since the beginning of the year, an increasing number of Google employees have expressed discontent with the organization’s politics.

“Sergey [Brin] did a 180 politically, Sundar stood behind Trump at his inauguration,” the current Google employee said. “Now one of the only remaining figures at the company willing to speak out on moral issues is gone. Who knows what we will sell out next?”