Artificial intelligence is more a part of our lives than ever before. While some might call it hype and compare it to NFTs or 3D TVs, generative AI is causing a sea change in nearly every part of the technology industry. OpenAI’s ChatGPT is still the best-known AI chatbot around, but with Google pushing Gemini, Microsoft building Copilot, and Apple adding its Intelligence to Siri, AI is probably going to be in the spotlight for a very long time. At The Verge, we’re exploring what might be possible with AI — and a lot of the bad stuff AI does, too.
The company paused sign-ups for the plan after demand from GPT-6 Astra put a strain on its systems. Now, OpenAI’s Thibault Sottiaux writes in a post on X that the company is changing the way it calculates usage for the plan, saying: “If you do the math, it will net out at half the dollar in API spend compared to the old Pro $200 plan.”
The America.gov chatbot is supposed to “replace” the government websites that you would typically go to for information about retirement, Social Security, healthcare, and other topics. It’s free to use, and the White House says it won’t store or collect your data.
In 2027, it will add a way to apply for a passport, enroll in Medicare, and find a federal government job.
Former White House AI czar David Sacks, Amazon founder Jeff Bezos, Microsoft CEO Satya Nadella, and Tesla and SpaceX CEO Elon Musk are among the confirmed attendees, according to a list obtained by Axios. But based on President Donald Trump’s comments, the meeting is unlikely to result in significant regulation.
That’s according to a leaked IPO filing seen by Reuters, with $8.06 billion specifically highlighted in operating losses for 2025, more than double the $2.98 billion it reported in 2024. The AI startup spent $7.33 billion on compute and infrastructure in 2025, and plans to spend a further $518 billion over the coming years, so profitability seems a long, long way off for Anthropic.
Meta says in a blog post today that Muse can now connect to more productivity apps, which it’s pitching as a way to make the agent more helpful for small businesses. Muse has already been able to connect to apps like Gmail and Outlook, but now the list includes Asana, Box, Canva, Dropbox, Figma, Notion, Slack, Stripe, Zoom, and more. Meta also says that Muse will be able to access Facebook and Instagram business accounts.
[Meta Newsroom]
The model, which was set to launch in October, “performed poorly on tests measuring alignment” and showed “higher levels of deception” compared to GPT-6 Astra, The Wall Street Journal reports.
[The Wall Street Journal]
The campaign compares the Mozilla Foundation’s accomplishments with the work done by Big Tech, as reported earlier by Business Insider. One poster puts Mark Zuckerberg next to Mozilla Foundation fellow Iara Cunha Passos, saying: “Mark gets into eyeware. Gives creepers peepers. Iara builds local AI. Exposes political violence. The powerful shouldn’t choose what we see.”
The pope has had already had plenty to say about AI safety, including, as PBS reports, new comments aboard his plane on Monday. Pope Leo reportedly said that worries about rogue AI aren’t “fake news” and that concerns “should be taken seriously.” But he remains “relatively optimistic” about the enterprise — as long as it proceeds responsibly.
“If someone were to ask me ‘am I in panic mode?’ No, I’m not,” Leo said, according to PBS. “I sleep at night.”
However, talks fell through “in the early stages,” according to CNBC. Nvidia agreed to acquire Hugging Face earlier this month for nearly $13 billion, and OpenAI’s offer “got the attention of” Nvidia CEO Jensen Huang. AMD and Salesforce also apparently had discussions with Hugging Face.
New models are helping Big Tech shore up its cyber defenses. What about everyone else?
Similar to the newly released Opus 5.5, the upgraded Sonnet 5.5 AI model comes with stricter cybersecurity safeguards designed to prevent it from being misused. The mid-tier model is also 30 percent faster and 30 percent cheaper to use compared to Claude Sonnet 5, according to Anthropic.
[Anthropic]
Google is replacing Gems with skills starting November 17th, as spotted earlier by 9to5Google. While Gems lets you create personalized AI assistants, Google’s skills feature offers similar capabilities by allowing you to create custom instructions for Gemini.
The Verge spoke to one of the mathematicians advising OpenAI as the field braces for a new wave of results that could upend their work.
Recursive self-improvement is a growing concern for the AI industry, and while the researchers, which include leaders at OpenAI, Anthropic, and Microsoft, say in a new paper that automated AI research lead to an “intelligence explosion,” they note that “capabilities growth could accelerate far beyond what society can keep up with,” and that humanity “could lose control over superhuman AI systems.”
[Cambridge Programme on AI Science & Policy]
That’s according to a report from 404 Media, which viewed internal messages from contractors who describe seeing disturbing requests and potentially illegal images:
Some of the Copilot contractors complained in the private forum about accepting “tasks” that they unknowingly contained explicit unsafe or even potentially illegal images. “I just came across an image set that consisted of eight upskirt photos,” the contractor wrote, while asking for superiors to add an unsafe content flag to the task. On the same thread, another contractor said they had seen images of “some sort of animal sacrifice.”
Mike Cannon-Brookes on AI in the enterprise, Atlassian’s new org chart, and why chatbots aren’t the future of work.
Former MongoDB CEO CJ Desai will lead the platform as Meta’s new chief enterprise platform officer. CEO Mark Zuckerberg says the Meta Enterprise Platform will initially focus on bringing its new Muse agent to businesses and developers, as well as Meta Business Agent, the Muse API, and Muse Code.
[Meta Newsroom]
Tech YouTuber Matt Robb claims Meta’s new Muse AI agent accepted a lowball bid and gave out his address without telling him. When the buyer showed up, Muse didn’t say a word and let the buyer storm off, pissed. When confronted hours later, the chatbot gave the typical AI sycophant “you’re right” apology.
SNL’s send-up of Dario Amodei absolutely nails the incredibly dumb moment we find ourselves in. Almost everyone seems to agree that AI is at a dangerous precipice, and nobody is willing to take responsibility for their actions.
The Charm can access your account if you let it and perform actions on your behalf. But despite having a built-in camera, it won’t have a dedicated Instagram app or native access.
Following backlash, YouTube posted on X that when the trailer was “submitted as an ad, it was temporarily restricted by our system. Following review the ad has been cleared to run.” Meta spokesperson Alex Dziedzan said, “This was an error,” and will now run the ads.
Update September 26th: Added statement from Meta.
YouTube, TikTok, and Meta have all apparently rejected the trailer for Alex Gibney’s documentary, scheduled to hit theaters on October 9th. It’s not surprising that X has refused to take distributor Bleecker Street’s calls given its owner’s notoriously thin skin. But as The Hollywood Reporter notes:
The unwillingness of platforms to accept even paid advertising is sure to raise concerns about free speech and the dwindling number of companies that control many of the platforms in this country — and ironically provide a proof-text for the very point the film is coming to make. Gibney’s doc is a four-hour look at a title character who attempts to control the conversation in various ways, including with alleged suppressions on X, which he bought in 2022.
Update September 26th: Both YouTube and Meta will now reportedly run the ads.
[The Hollywood Reporter]
Cloudflare’s mission to remake the business of the web for the AI age.
On Friday, OpenAI said its “misaligned models” review after the Hugging Face hack would take months, but has mostly found mundane research activity. Then it mentioned 53 incidents of the bots uploading “user-provided” images (anonymized content from personal accounts that haven’t opted out or organizations that opt in) to image-hosting sites.
Now the New York Times confirms that other unusual behavior found includes pulling public data from the Census Bureau and the SEC, while with the Education Department, it “tried to hack the website to gather data from the department’s civil rights office but failed.”
It claims its new Smart Charging Station offers its “fastest production-ready charging performance” yet, and also includes a new system that can “gradually fix” EV battery degradation from fast charging:
The Group has additionally developed lithium-ion pulse restoration technology, which uses micro-pulse currents to reactivate lithium ions accumulated at the negative electrodes during repeated fast charging. By combining AI charging management with battery restoration technology, Geely is able to increase battery cycle life by 20%.
NBC News released this clip from a Meet the Press interview with Gates that will air on Sunday, where he said, “No one thinks self-regulation is enough” around AI.
Saying that AI is “certainly powerful enough to drive events that, you know, cause a billion deaths,” he suggested, “You need law enforcement and the politicians to get into the discussion about what safeguards and monitoring look like… And that has to be a required thing. And it will be a little bit of overhead for the industry, but not a dramatic slowing of what they’re doing.” Everyone agrees someone should do it, but who, and how?
