Moonshot AI's Kimi upends conventional wisdom on U.S. lead over China People walk past the Moonshot booth promoting its Kimi K3 AI model, during the World Artificial Intelligence Conference in Shanghai on Friday. At an event in Beijing earlier this year, some of China's top artificial intelligence leaders warned that the country remained meaningfully behind the U.S. in developing cutting-edge AI models, with one executive arguing that "the gap may actually be widening." Leading U.S. firms also appeared confident they were significantly ahead. As recently as last week, one executive at Anthropic, who spoke on condition of anonymity, mused that the Claude maker's technology was roughly six to 12 months ahead of Chinese rivals. The Chinese AI lab released Kimi K3, a more advanced open-weight model that it said outperforms all rivals except for Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 on overall capability.

"Context bombing" tricks malicious AI agents into shutting down before they can do harm. Prompt injections, the malicious commands attackers embed into content to entice large language models to follow them, have been attackers' go-to tool for turning AI platforms against their users. A well-phrased command sneaked into an email or calendar invitation is often all it takes to cause the LLM to exfiltrate sensitive data or follow other harmful actions. Now, defenders are embracing the prompt injection, too. Researchers from Tracebit on Monday said they found that placing prompt injections alongside passwords, cryptographic keys, and other secrets stored on Amazon Web Services was often all that was needed to shut down attacks from AI hacking agents.

Follow this section to personalize your feed and get instant alerts. Follow Go to your personalized feed WHY FOLLOW? Smart Alerts: Get notified about major news as it happens. Follow this tag to personalize your feed and get instant alerts. Follow Go to your personalized feed WHY FOLLOW?

On today's, we unpack OpenAI's ongoing drama, both legal and reputational, and whether these developments could further hurt the company--particularly in its fight against Anthropic. This week on, the team discusses the details behind Apple suing OpenAI over alleged stolen hardware secrets. And the company's headaches don't stop there--a group of OpenAI employees just started a super PAC to advocate for stronger AI guardrails. Plus, New York's first-in-the-nation data center moratorium draws Donald Trump's ire, DOGE stonewalls FOIA requests on its AI use at HUD, and WIRED's Emily Mullin explains the cyclosporiasis outbreak spreading across more than 30 states. DOGE Used AI for Housing Policy. The Government Won't Say How Write to us at [email protected] . You can always listen to this week's podcast through the audio player on this page, but if you want to subscribe for free to get every episode, here's how: If you're on an iPhone or iPad, open the app called Podcasts, or just tap this link . Last Friday, Apple sued the company basically alleging that OpenAI has been stealing confidential hardware secrets. And now this week, WIRED learned that some OpenAI employees are launching a super PAC to push for AI guardrails. We'll get into all those details and whether these developments could further hurt OpenAI, particularly in its fight against Anthropic. We'll dive into what that means exactly and whether this move could pave the way for other states to follow suit. And we're going to check in on the disease sweeping the nation, cyclosporiasis, which is causing turbo diarrhea across the country . So basically what happened was last Friday, Apple filed a lawsuit against OpenAI for allegedly stealing information like unreleased iPhone parts and prototypes, confidential designs, documents about secret projects. That accusation sucks for OpenAI. It's pretty intense, but it gets really messy because Apple is basically saying that this theft, this alleged theft, mostly happened through former employees.

This feels like the epitome of a distracting side quest. This past spring, former OpenAI executive Fidji Simo warned employees the company was at risk of missing its moment because it was distracted by side quests, referring to costly software projects like Sora, which OpenAI shut down in April. This week, 's Mark Gurman shared an update on the company's nascent hardware efforts, writing that OpenAI's debut first-party device would be a humanlike rechargeable smart speaker . For an unprofitable company, this feels like the epitome of what Simo hoped to avoid. In terms of the trend line, it's been slowing for quite some time, and we're expecting declines in the upcoming year, says Jitesh Ubrani, the director of worldwide device trackers at market research firm IDC .

If you like putting Claude's AI agents to work on personal chores but are concerned about giving them access to your credentials, there's a new solution available. Users can now authorize Claude to complete real-world tasks like booking travel and managing accounts securely with credentials injected directly to the target system on their behalf, 1Password explains. Security has emerged as a big issue when using agents to complete tasks on your behalf. Sharing credentials directly with an agent exposes them to the model, its memory and the systems behind it, which can pose a risk if the model gets hacked . To solve that problem, 1Password built a zero-exposure security framework that allows agents to use stored credentials from your 1Password vault without them reaching the model.

It also alleges that Google "downloaded web scrapes of virtually the entire internet, including from known pirate sources and from behind legitimate paywalls". It further alleges that Google copied those works without permission to train its AI models and continues to do so, despite those uses allegedly falling outside the scope of existing agreements. The suit claims the company was fully aware of the legal risks, alleging that internal documents warned using books to train AI models was "highly problematic for Google," and could lead to as much as $100bn in fines. "At no point did Google inform authors and publishers that Google was copying their works as source material to develop and train AI models," the suit alleges. "It's an interesting issue that has a lot of complex dimensions, in no small part because it can be hard to prove what was or wasn't in a training corpus."

The new collaboration with keyboard maker Work Louder is available to order today. The first piece of OpenAI hardware is here, and no, it has nothing to do with Jony Ive . It's called the Codex Micro, and it's a keypad created in collaboration with keyboard maker Work Louder, specifically for controlling OpenAI's agentic coding feature Codex . The two companies started teasing the accessory in June, and the Codex Micro is available to order now for $230, ahead of the Jony Ive-designed smart speaker OpenAI is rumored to be introducing later this year. If you've seen Work Louder's Creator Micro 2, or the Framer Micro the company created in collaboration with the website builder Framer, you have a good sense of the fit and finish of the Codex Micro.

OpenAI employees have donated more than $215,000 to a political effort opposing Leading the Future, a group backed by the company's president, Greg Brockman. A group of rank-and-file OpenAI employees have donated more than $215,000 to a super PAC pushing for stricter regulations on frontier AI labs. Guardrails Alliance, which launched last month with $5 million in total initial funding, bills itself as a populist effort supported by tech workers, labor unions, and other groups. It's aiming to be a counterweight to Leading the Future, a pro-AI industry super PAC bankrolled with more than $100 million from technology industry leaders, including OpenAI president and cofounder Greg Brockman. Seven current OpenAI employees have donated to Guardrails Alliance, as well as one former employee, WIRED has learned.

OpenAI's first device will reportedly be a'humanlike' rechargeable speaker OpenAI's first device will reportedly be a'humanlike' rechargeable speaker It will be powered by a more advanced version of the model used by ChatGPT's new voice mode, Bloomberg says. It has long been reported that OpenAI's first hardware product will be an AI-powered speaker . Now, a new Bloomberg report reveals more details about the upcoming device, which will apparently be rechargeable and can easily be carried from one room to another. You're supposed to be able to take it to the kitchen, for instance, to help you with cooking instructions and then carry it back to your living room or bedroom. The publication says the speaker is meant to be a humanlike AI companion that can talk to you naturally in the way ChatGPT can, which can also control your smart home devices and play audio.