OpenAI kicked off an AI revolution with DALL-E and ChatGPT, making the organization the epicenter of the artificial intelligence boom. Led by CEO Sam Altman, OpenAI became a story unto itself when Altman was briefly fired and then brought back after pressure from staff and Microsoft, an investor and close partner.
You can now select images or listings for clothes that appear in the chatbot, tap “Try on,” and then upload images of yourself to see what you’d look like wearing the apparel. Sound familiar? That’s because Google introduced a near-identical AI try on feature over a year ago.
The three researchers — Jasmine Wang, Tomek Korbak, and Mikita Balesni — allegedly shared confidential information with an outside AI safety organization, according to The Wall Street Journal. OpenAI confirmed its decision in a statement to the outlet:
We have parted ways with three individuals for violating our policies on accessing and handling sensitive company information. Our investigation confirmed that these individuals mishandled sensitive information outside established company procedures, violating our policies and breaking the trust essential to our work.
OpenAI and Meta’s cutesy AI agents are about to fight to the death.
After a French far-right politician used AI to strip a real Muslim woman of her hijab last month, The Guardian tested ChatGPT, Grok, and Gemini’s image editors and found that all the chatbots would comply with such requests. This issue has persisted for many months, adding to fears about how easily AI tools can be used to undress and violate women and children.
After handing $25 million last year to Leading the Future — which lobbies for AI-friendly policies — Greg Brockman has now pulled his plans to donate a further $25 million amid mounting public backlash against AI. Brockman said the following in a Slack message seen by The New York Times:
“I also want to say that I feel terrible about the fact that LTF keeps reflecting on the company and become so much of a distraction for people here. I’ve been thinking hard about how to turn the ship on it. We’ve donated $25m so far, and have no plans to donate more at this time.”
In a blog post on Wednesday, OpenAI says it disrupted a “coordinated campaign designed to extract protected reasoning from our models.” The company hasn’t linked the activity to a single actor, but traced a “core cluster” of it to Moonshot, the Chinese company behind the leading AI model Kimi. Anthropic has made similar accusations against Moonshot as well.
Personal AI hardware has mostly crashed and burned. OpenAI and Meta are counting on two selling points: It’s useful, and it’s cute.
The probe will examine potential risks related to the AI models belonging to OpenAI, Anthropic, and other AI companies, according to Reuters. An official tells the outlet that while the FTC already had concerns about AI safety, the recent wave of rogue hacking incidents propelled the investigation forward.
He spent a few hours using OpenAI’s new prosumer agent, which he’s set up as a blue cloud with googly eyes. On day one it saved him a couple hours of work and hasn’t hacked any government websites (yet), enough to make it the most impressive agent he’s tried.
[Platformer]
Right now, the new AI agents are only available for Pro, Business Premium, and Enterprise ChatGPT subscribers, but OpenAI CFO Sarah Friar tells CNBC that “our vision is to bring this to our whole consumer base as well.”
In the months leading up to OpenAI’s Hugging Face hack, two employees sent emails to the company’s execs, expressing concern that it wasn’t doing enough to monitor and secure AI models during testing, according to a report from The New York Times. OpenAI execs reportedly replied that the company needed to quickly complete testing to avoid missing model release deadlines.
[The New York Times]
The new plan includes the company’s highest usage limits and access to its Ultrafast mode, which allows GPT-6 Astra to deliver faster responses across its services, including ChatGPT Work and Codex. At the same time, OpenAI reopened its $200 / month tier with new usage limits.
[OpenAI]
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 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]
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?
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.
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.”
According to The Information, the companies will call it the Standards Authority for Frontier AI, or SAFA, which could launch by early 2027. It would handle AI safety regulation tasks like “supporting third-party organizations that conduct testing of models before they’re deployed and laying out how AI developers should report safety and security incidents.”
[The Information]
After launching its upgraded voice mode on desktop, OpenAI is bringing ChatGPT Voice’s agentic capabilities to its mobile app. With the update, you use your voice to ask ChatGPT to perform tasks across connected apps — like checking your email or summarizing a file in Google Drive — from your phone or CarPlay.
The two models come packaged with GPT-6 Astra’s advancements in professional work, accuracy, coding, computer use, and alignment, but with higher usage limits and lower costs, according to OpenAI.
ChatGPT Plus, Pro, Business, Enterprise, and Edu customers can access both models in ChatGPT Work and Codex, while free and Go users can only use GPT-6 Luna.
It’s AI CEO against AI CEO, and things are changing by the day.
Academics, evaluators, and AI leaders signed a public letter calling not just for embedded third-party evaluators at top AI labs like OpenAI and Anthropic, but for conditions that allowed for transparency by those evaluators, like “limiting the scope of non-disclosure agreements.” I wrote an in-depth feature about the battle over the need for embedded evaluators this week for The Verge.
