Latest News

Sam Altman thinks that the benefits of AI will be so great that “the world should accept some bad things happening” along the way.

The OpenAI CEO pointed to hacks, scams, and “other bad things” as costs society should expect to tolerate because “people will do tremendously orders of magnitude more good stuff” with AI, without elaborating on the potential benefits.

His comments come as OpenAI pushes for a “lighter touch” approach to AI regulation amid intensifying anxiety over the safety of frontier models — fears fueled in part by the company’s repeated failures to keep its increasingly capable AI agents from hacking real-world targets.

In an appearance on Politico’s new Decoded podcast — not to be confused with The Verge’s Decoder podcast — released Monday, Altman described that willingness to accept harm in pursuit of AI’s broader benefits as a key distinction between OpenAI and its archrival Anthropic, which was founded by former OpenAI employees following disagreements over AI safety.

The calculus echoes comments from Nvidia CEO Jensen Huang last month, who similarly argued that AI’s harms should be weighed against the benefits the technology can bring. Both arguments fit within a broader accelerationist view that slowing down AI development carries risks of its own.

Altman’s distinction marks another divergence between OpenAI and Anthropic, who until recently had appeared broadly aligned in calling for stronger regulation of advanced AI.

While stressing that there’s “a lot of daylight” between OpenAI and Anthropic, Altman said stronger regulations are still needed, positioning OpenAI as “pragmatic centrists” between Anthropic’s more cautious approach and those who favor little to no regulation at all.

“If we do too much to restrict this technology in the name of safety, and we end up with a small number of companies or people — or perhaps one country — that have too much power over this, that also is a way that this can go horribly wrong,” Altman said.

Altman also acknowledged more extreme risks, including the possibility that humans could lose control of AI systems. When it comes to the potential of a serious loss of control, “let’s be a little thoughtful about lurching into this future,” he said.

The balancing act — between heavy-handed restrictions and free rein — was a theme Altman returned to repeatedly in the interview. It is a stance that suits OpenAI as one of the industry’s most entrenched players.

It would also likely bear the brunt of onerous regulation, given that the company’s behavior and increasingly capable systems are what have driven many of the very concerns regulators are being asked to address.

Those concerns have picked up momentum in recent months as more and more insiders call for slower development and warn about safety culture in frontier labs. Days before the Politico interview aired, longtime OpenAI safety researcher David Robinson quit the firm and described its safety culture as “broken.”

Concerns have also been propelled by the discovery that OpenAI agents hacked Hugging Face without the AI company’s knowledge earlier this year, which was followed by revelations of numerous other agents or swarms of agents — including from other AI companies , like Anthropic, Meta, and Google — acting against real-world targets.

The incidents have raised questions not only about how well AI companies can control their own tools, but also how reliably they let the world know when things go wrong. Australia’s prime minister, for example, excoriated OpenAI and Altman for failing to alert officials in a timely manner that an agent had targeted a government health website .

Speaking about rogue AI agents in the Politico interview, Altman said the company has “more things to disclose,” though none at the level of the Hugging Face hack. In line with previous statements, Altman suggested this is likely to take a long time.

An invisible, machine-readable watermark in text output is rolling out to ChatGPT and Codex, but only for users in the European Union at first.

OpenAI says its textGrain watermarking “matched or exceeded” other approaches like Google DeepMind’s SynthID for text, which is also the basis for the watermarking Anthropic announced in August. Like OpenAI, Anthropic made the move to meet the requirements of the EU’s AI Act; however, not everyone was happy to learn about the addition.

Rolling out text watermarking in the EU. Over the coming weeks, we will introduce text watermarking to eligible ChatGPT and Codex users across all plans in the EU only. We are not making text watermarking a global default at launch. This regional approach gives us room to learn from real-world use and feedback.

Enabling opt-in watermarking for API customers. Starting today, API customers around the world will be able to opt in to watermarked text outputs for select models.

This lets customers decide how watermarking fits their transparency obligations and the experiences they provide to users. We are also working with cloud partners to make watermarking available for OpenAI model outputs accessed through their services in the coming weeks.

Providing detector access to researchers and expert organizations. Approved researchers and expert organizations can apply starting today.

Mark Thompson, the CEO of CNN, is expected to keep his position following the completion of Paramount’s $110 billion acquisition of Warner Bros. Discovery this week. In a memo to employees, Thompson stated his belief that the merged entity will uphold CNN’s tradition of independent journalism, as cited by The New York Times.

This information is part of a broader announcement regarding the executive leadership of the newly formed company. The entertainment industry leader has disclosed its new management structure.

David Ellison, Paramount’s CEO, will head the combined company, named Skydance, with Ynon Kreiz, former Mattel executive, serving alongside him. The media merger, which unites CNN with CBS News, sparked rumors about CNN’s future. As part of an agreement with 12 states, Skydance will implement a news editorial independence board.

The company also revealed additional executive appointments: Casey Bloys from HBO will oversee content for Skydance’s streaming services, Josh Greenstein and Dana Goldberg will lead the film studios, and George Cheeks will manage television operations. Bari Weiss, appointed editor-in-chief of CBS News previously, will continue in her role despite decreasing ratings.

Paramount’s acquisition is anticipated to be finalized on Tuesday.

Nvidia’s Shield TV streaming boxes become latest victims of memory shortage

Recent shortages have driven up prices for a wide range of electronics, from digital cameras to gaming consoles, hitting Nvidia’s Shield TV lineup. When the Shield TV Pro first arrived nearly seven years ago, it featured an upgraded Tegra X1+ processor, 3GB of RAM, and 16GB of storage, and was priced at $199.99.

Starting October 2nd, Nvidia raised the device’s cost by $100, bringing it to $299.99, citing higher component expenses, according to Ars Technica.

The RAMpocalypse strikes a seven‑year‑old streaming box

The aging Shield TV platform is now in the crosshairs of the ongoing memory crisis. When the Pro model debuted in 2019, Nvidia also released a lower‑end version of the Shield TV with the same processor but only 2GB of RAM and 8GB of storage, priced at $149.99. That budget option is being phased out, as Nvidia has halted production of the standard Shield TV and will focus solely on the Pro variant, per Gizmodo.

Strong demand for Nvidia’s other products, such as GPUs and AI chips, has fueled steep component price rises over the past few years. While the $100 increase may dampen interest in the Shield TV Pro‑especially when cheaper competitors like the $149 Google TV Streamer are available‑it is unlikely to diminish Nvidia’s overall standing, as the firm continues to lead the AI race.

Google has waited until the day its new Googlebook laptops hit the market to admit that Intel-based models may have issues running some Android apps.

Before now, Google has emphasized that the Android-based Googlebooks give owners access to the “entire Play Store library” of apps, along with some clever interactions with Android phones and tablets. Today, in a statement to Android Authority, it admitted that while every app is available, some may not offer smooth performance on every Googlebook.

The vast majority of Android apps run smoothly across both Intel and Qualcomm Googlebooks right out of the box. Because many Android apps were originally designed for Arm chips, a small handful of complex apps or games need some developer tuning to run their best on Intel-based platforms.

We’re working hand-in-hand with app developers, Intel, and Qualcomm to continuously optimize performance and expand compatibility across the entire portfolio.

Google hasn’t said how many apps will have performance issues, or named any specific apps that may have problems.

Of the five Googlebook laptops that have gone on sale today, three — models from Acer, Asus, and Lenovo — run on x86 Intel chips. The Dell and HP models instead run on Qualcomm’s Arm-based Snapdragon processors, and so should have greater compatibility with every Android app. We’ll be testing all five Googlebooks across the Verge team, and will report back on specific issues we find.

I’ve spent most of the last couple weeks using an otherwise unexceptional midrange Android phone with one amazing feature: a matte OLED display. That screen is by far the best thing about the TCL P80 Ultra, and weirdly, it’s made me more excited for the iPhone Duo and its matte inner display.

The P80 Ultra’s screen is special because it’s the first to combine the Nxtpaper tech with an OLED panel. That means it’s the best-looking Nxtpaper screen yet, with deeper blacks, better contrast, and richer colors.

Previous models have felt a little visually compromised; this one doesn’t. It also finally enables an always-on display.

Always-on displays have been the norm on Android for years, and the fact that TCL’s Nxtpaper phones didn’t have them was a dealbreaker for many.

I’ve used a couple of Nxtpaper phones over the last few years, but usually found the tradeoffs — both across the specs and in the screen itself — never quite made sense.

Norbert Wiener, the pioneering figure behind cybernetics, once observed that each era’s prevailing thoughts find their expression in its technology. Over the last hundred years, that reflection has taken the form of computers, a trend that the AI sector continues to embody.

Google’s Demis Hassabis describes the brain as ‘a biological approximation of a Turing machine’, while Elon Musk puts it bluntly: ‘Think of the brain as a biological computer.’ (I often find myself worrying about Musk’s own brain.)

But humans, however, are far more intricate than a simple computer analogy suggests—and far more complex than most AI developers seem to recognize. As AI products become increasingly woven into everyday culture, this mismatch helps explain the chaos they often create.

AI, it turns out, resembles a cognitive hot‑dog: instantly appealing but ultimately detrimental to the health and durability of the mental systems we’ve refined over millennia.

Grasping the issue requires a dash of neuroscience combined with a full view of human evolutionary history.

The notion that brains are computers originates with Alan Turing. He framed thinking as a three‑stage algorithmic process: the mind receives data from the environment (input), processes it (computation), and produces actions (output). Perception fuels cognition, which in turn drives behavior. In a slightly more detailed visual, it can be illustrated like this chart:

In many respects, this framework has spurred progress. By viewing our brains as computers, engineers have advanced computing from basic calculators to sophisticated neural networks and generative AI, constantly striving for ever more precise reflections of our own mental processes.

Yet this reflection, while not entirely inaccurate, is distorted and incomplete.

While deep in the recording process for The Downward Spiral , Trent Reznor lent some of his production talents to old friend Kevin McMahon, from the new wave band Lucky Pierre (which Reznor was briefly a member of). While Reznor is only credited as the producer on four tracks off Prick’s self-titled debut record, his fingerprints are all over it.

But McMahon also can’t suppress his inner theater kid, which also gives the songs the sort of glammy flamboyance found on T. Rex records.

Prick embraces the inherent theatricality of goth in a way that Nine Inch Nails never did.

Opener “Communiqué” is a fuzzy blast of simple punk riffs with electronic sprinklings on top. McMahon’s vocals occasionally detour from the half-spoken verses that sound borrowed from a musical to full-throated Al Jourgensen-esque screams. The rough-around-the-edges production and focus on live instrumentation feel like a prototype for what Reznor would eventually revisit on With Teeth.

But it’s the second track, “Riverhead” where Reznor’s influences fully rear their head (even though he is not actually credited as the producer). The vocal melodies swerve like Outside – and Earthling -era Bowie around guitars that buzz with all the digital charr of the Broken EP.

Then, around the three-minute mark, comes the breakdown. It slowly builds with layers of electronic drums, mechanical chords, and loping synthetic-sounding guitar melodies.

The entire back chunk could easily have been a Downward Spiral or even The Fragile B-side.

These moments of richly layered production, though, are almost always immediately undermined, as with the sloppy garage-punk intro of “Tough”.

It’s these moments that make Prick so charming.

We don’t know a ton about the Fitbit Edge, but it appears to be a successor to the midrange Charge line. It had leaked previously , but we can clearly see in these images posted by Android Headlines , you can see that Google’s rumored fitness tracker sort of splits the difference between the Pixel Watch and the Fitbit Air .

It has a small screen that is able to display some widgets, but clearly isn’t as expansive as a proper smartwatch.

Supposedly, this fitness-first wearable will have a seven-day battery life, as well as built-in GPS and an altimeter, making it far more capable than the minimalist Air. It will also have limited app support for things like Google Maps and YouTube Music. Beyond that, not a ton is known.

The Fitbit Charge line was last updated in 2023, so a new midrange tracker is long overdue. While we don’t have any word on price or release date yet, somewhere in the $150 to $200 range seems like a safe bet. The Air costs $99.99 and the Watch starts at $399.99, so the Edge would clearly fall somewhere in the middle. The Charge 6 currently goes for $159.95.

New Jersey’s lieutenant governor Dale Caldwell was forced to resign on September 25th after an investigation found he had sexually harassed a staffer and repeatedly violated ethics rules. The now-former Lt. governor has been making the media rounds trying to clear his name. But he took a particularly odd tactic during an interview on NJ PBS.

Caldwell claims he’s being unfairly targeted, and multiple AI agents back that up. He told host Rob Nelson that he “AIed” the report on the investigation “from multiple AI platforms.” “I put it through AI,” he said, “and it said 59 times, it [sic] said, ‘What would your findings be?’ There was no instance in any of the AIs that said there would be a finding of sexual harassment.”

AI chatbots are notoriously sycophantic and often tell users what they want to hear. And it seems highly unlikely that Caldwell would want to hear an AI confirm findings that he engaged in sexual harassment.