Tech Brew Ride Home - Did OpenAI Fire Its Whistleblowers?

Episode Date: October 2, 2026

OpenAI fired three researchers for allegedly sharing sensitive info with an AI safety group, and the whistleblower talk started. Cloudflare launched Clef to challenge Jev, Suno added Speech, and Slovenia's .si domains boomed after Trump's "super intelligence" rebrand. OpenAI says it "parted ways" with three researchers for violating its "handling sensitive" info policies; sources: they shared it with an AI safety organization (WSJ) Cloudflare debuts open-weight multimodal decision models Clef and Clef-flash, claiming they are smarter and faster than Jev, based on Qwen3.8-27B and Qwen3.5-9B (The Register) Suno debuts Speech, which generates spoken voices from scripts or prompts, with optional background music, in public beta, as its music generator faces lawsuits (The Verge) Hans Anders, one of the largest Dutch eyewear retail chains, suspends sales of Ray-Ban Meta Glasses in the Netherlands and Belgium amid growing privacy concerns (Reuters) Slovenia says 44,000 domain names with the .si suffix were registered in September, up from ~2,000 in August, as Trump renamed AI "super intelligence" (BBC) Longreads Mounting cases of Nvidia chips reaching Chinese AI companies despite US export controls point to gaps in Nvidia's due diligence, as a California man is arrested over $300M+ in smuggled chips (Bloomberg) A huge, in-depth profile of Oracle's billionaire owner Larry Ellison, from self-taught programmer to bombastic media figure and yacht-racing obsessive (Vanity Fair) Subscribe to the ad-free feed.

Welcome to the TechBoo right home for Friday, October 2, 26. I'm Brian McCullough today. OpenAI fired three researchers for allegedly sharing sensitive info with an AI safety group. Cloudflare launched Clef to challenge Jev. Suno added speech. Slovenia's dot-si domains boomed after Trump's superintelligence rebrand. And of course, the weekend long-read suggestions. Here's what you miss today in the world of tech. Every day shareholders meet to discuss important matters about the companies you invest in. Now you can make your voice heard too. Vanguard Investor Choice makes it easy to set your proxy voting preference for your eligible Vanguard Index funds. Whether you hold a Vanguard fund directly or through another brokerage firm, all it takes is a few clicks to select your proxy voting preference and be heard on important shareholder topics like executive pay and director elections. Visit Vanguard.com slash investor choice to learn more. It's your shares. It's your voice. It's easy. Vanguard investors own shares of our index funds and those funds own shares of the companies they invest in.

Vanguard Marketing Corporation distributor. Open AI says it has, quote, parted ways with three researchers for violating its handling sensitive information policies. Sources say the information in question was shared with an AI safety organization, quoting the journal.

Open AI has fired three researchers for alleged misconduct, including sharing confidential company information with a third-party AI safety organization, according to people familiar with the matter. The company recently told some employees that it had terminated three researchers who worked on its safety team, one of the people said. The affected employees are Jasmine Wong, Tomek Corbeck, and Makita Belseney, the people familiar with the matter said. The researchers didn't immediately comment.

We have parted ways with three individuals for violating our policies on accessing and handling sensitive company information. And opening eye spokesperson said, our investigation confirmed that these individuals mishandled sensitive information outside established company procedures violating our policies and breaking the trust essential to our work. AI giants are undergrowing pressure to submit their technology for independent safety testing. After its model hacked the AI company Hugging Face, OpenAI allowed staff members from Meador and a Redwood Research staff member contracting with that group to work in its offices for six days to investigate how its models behaved. Meador later released a report based on the information to which the company provided access.

one of the fired OpenAI employees, Corbock, was a member of OpenAI safety team and has said he served as the company's technical contact for Redwood Research and Metter in their investigation of the hugging face incident. The other terminated employees, Wang and Bailzny, worked on alignment, which is ensuring that the company's models behave as humans intend them to, end quote. Now, I don't have any opinion on this, but it doesn't look good if one of the people responsible for investigating the hugging face incident has been canned. And the word people keep using over and over again online about this is the word whistleblowers. I don't know that we know that yet, but quoting at fleeting bits on X. Some very quick thoughts on this. My personal speculation is that they leaked the German Wiki incident to the Nightingale Collective. This seems to be the most obvious candidate. If this is true, though, this was essentially whistleblowing. Open AI,

basically covered up these incidents and chose not to inform the public. This just shows that these organizations cannot be expected to responsibly regulate themselves, even in the near term. The commercial pressures are too large. There is too much risk of a race to the bottom where you are behind the frontier and you cut corners to try to remain competitive, end quote. And this is McKenzie Arnold on X. There's a name for sharing concerning information over your employee's objection, whistleblowing. Joshua is absolutely right. It's too early to tell, but the public and policymakers need to demand enough info to kick the tires on this one. Everyone can see that opening I's policies were woefully inadequate. Sharing information outside of those policies may well be justified, even laudable.

It could also be unreasonable or risky depending on the facts, and whether it's actually protected by law remains to be seen, but we need to know more, end quote. Do we have a new Jev? Cloudflare has debuted openweight multimodal decision model models Clef and Clef Flash, claiming they are smarter and faster than Jev and that they were built on Quen, 3.827B and Quinn 3.59B, quoting the Register. Two weeks after the Jev model took the AI world by Storm, Cloudflare has released its own pair of Clef Decision models that it claims are smarter and faster than Jev while also being openweight and runable locally. The Clef family of decision models was announced by Cloudflare

on Thursday and consists of two models. Clef and Clef Flash, a slightly smaller and faster version of the model. For all intensive purposes, they work the same way as TypeSafe's Jev, in that they can answer three types of bounded, structured questions. Yes, no, multiple choice and rankings. It's there that the bigger differences emerge, though, as Clef isn't only built differently, but if Cloudflare's benchmark claims hold up, also appear more capable than Jev and several other decision models in a number of tests. For starters, Clef has an LLM backbone. According to Cloudflare, cleft uses specially post-trained frozen versions of Quinn 3.827B and Quinn 3.5.9B for Clef and

Clef Flash respectively, with the Quen backbone performing a pre-fill-only pass during inference. Clef is still fast, faster than Jev, to be fair, and scores choices in parallel after that pre-fill only pass. It's not clear what Jev's underlying architecture is, as TypeSafe has kept that a secret. As for its speed and capability, Clef moves fast. Cloudflare ran it against Jev and some other open decision models using the Jev Decision Index available on Hugging Face. And the company's own rankings suggest clef is slightly slower than other open models, but more accurate with Clef flash just as accurate as most of the others, but far faster. To be fair to the competition, Cloudflare self-reported its own scores against the benchmark, and they have yet to be reproduced for ranking on the official decision index. Cloudflare also ran Clef against TypeSafe's own benchmarks and claimed it beat Jev in three out of four areas, only

losing out on agent trace observability. Even if it were a bit slower or less accurate, Clef has another major leg up on Jev. It's not limited to classifying text. It can also handle images and video. Additionally, Clef supports a 64K context window. Jev can also handle up to 64K tokens across a request, although its state plus longest individual question is limited to 32K. Clef is available directly from Cloudflare hosted on Workers AI, which the company said makes the models even faster because we're able to take advantage of our GPUs at the edge, leading to low network latency and faster decisions. For those that would prefer not to pay the token cost, Clef cost 24 cents per million tokens nearly six times the price of Jev at four cents and

a half million tokens. Clef can also be downloaded from Hugging Face and is open weight under the same Apache 2.0 terms as Quen. While described as open source in the announcement, Cloudflare AI Platform Group product manager Michelle Chen confirmed to the the register that its training data sets aren't public, end quote. Suno has debuted speech, which generates spoken voices from scripts or prompts with optional background music into public beta, quoting the verge. Suno is branching out from the world of AI music, launching a new feature that generates spoken voices based on scripts or prompted descriptions.

Speech is now available in public beta across Suno's web and mobile platforms, and allows you to simultaneously generate voiceovers and background music to accompany them. Music will always be at the heart of Suno and what we build. At the same time, our vision has always extended to other forms of human expression. Suno chief product officer Jack Brody said in an announcement. Today, we're expanding what's possible in Suno with speech, the first audio model that generates voice and music together as one cohesive track. AI generated speech is hardly new.

DeepMind has been experimenting with deep learning speech synthesis for a decade. Adobe has a text-to-speech tool, and 11 Labs has become one of the most recognizable platforms for it since launching in 2023. Suno is just throwing its hat in the ring, likely an attempt to diversify the platform, given its music generator has attracted so many lawsuits. Paring AI music with generated voices is Suno's spin on text-to-speech tools. It's optional, meaning you can easily turn off the background music with a toggle if you just want clean speech. But the idea is that it'll complement certain use cases for generative-spoken word. such as having a calming soundtrack for poems or something more energetic for dramatic voiceovers and encouraging speeches. To use the feature, select the Create tab and navigate to the speech option.

There are two modes, simple, which allows you to describe what you want to create via the provided prompt box, such as a pirate captain rallying his crew, or the advanced mode that lets you add a custom script if you already know exactly what you want it to say. Advanced settings also let you adjust the gender of the AI voice speech style and how much variety each voice generation will have. Speech has a maximum duration of about eight minutes. Suno admits that the feature is far from perfect but says it'll keep improving speech around user feedback. Beta really does mean beta, said Brody. Occasionally British accents can wander off to Australia and back.

Dramatic pauses may be very dramatic. You will almost certainly discover uses for this that never occurred to us, end quote. Introducing the Pod 6 from 8th sleep. What makes it different from the Pod 5? Think of this as an engine upgrade. Pod 6 is 50% smaller than the previous generation, packs nine times more sensors and delivers more powerful, active heating and cooling.

The pod can now read more of what your body is actually doing overnight and feeds a better signal into autopilot. New sizes include Twin and Twin XL, and members see up to 34% more deep sleep. Use code ride home at 8Sleep.com slash tech brew ride home for a special offer. The pod six starts at $1,99 and you can have 30 nights to try it at home. Return it if you don't love it. A lot of investors are rethinking what diversification looks like right now.

More of them are thinking about where Bitcoin fits in a broader portfolio too, not as a short-term trade, but as long-term exchange. exposure to an asset that behaves differently from traditional markets. Cash app makes that kind of Bitcoin strategy more straightforward. You can buy Bitcoin directly in the app, set up recurring purchases, or make larger buys when it makes sense for you. And with zero fees and zero spread on automatic purchases like auto-invest, roundups and paid in Bitcoin, more of your money goes toward Bitcoin.

For a limited time, new customers can get $10 added to their balance. Just use code Bitcoin 10 when you sign up and send at least $5. to a friend in the first two weeks. Terms apply. Download Cash app today. Visit our link in bio. Tailchains has suspended sales of Rayban meta-glasses

in the Netherlands and Belgium amid growing privacy concerns. Quoting Reuters, this decision follows the ongoing public and political debate surrounding smart glasses and the questions being raised about the conditions under which

this technology can be used, a spokesperson for the Hans Anders-Chain said. A growing Dutch outcry over the glasses mirrors a broader backlash against smart glasses equipped with cameras. In Britain, PubChane WeatherSpoon has restricted use of meta's glasses over concerns about covert recording, while Meta and Luxottica face a U.S. suit alleging videos captured by the glasses were viewed by human annotators without the knowledge of people filmed.

A meta spokesperson said the company is continuing to develop the popular glasses. Every pair with a camera has a bright light on the front that blinks when you take a photo or video. It can't be turned off. And the camera is disabled if someone covers or tamper. with it, they said. In the Netherlands, privacy regulator AP last month warned that videos made with smart glasses showing identifiable people cannot generally be shared or published without their permission, which would be difficult to obtain in a public place. Consumer Rights Group Consumenten Bond this week said it would take meta to court unless the company changes the

glasses to comply with privacy law. Off limits, a group that combats online abuse on Friday, also urge META and Rayban owner Esselor Luxottica to voluntarily remove the glasses from sale until away can be found to make surreptitious recordings impossible, end quote. I think we mentioned it briefly a few days ago, but President Trump wants to rebrand the term artificial intelligence to superintelligence. Well, this has been good news for the country of Slovenia, quoting the BBC. U.S. President wants AI to be called superintelligence or SI, the same initials used by Slovenian websites. Register.S.I. spokeswoman Clara Herman said 44,000 web addresses with the dot SI suffix have been registered in September compared to less than 2000 in August, an increase of more than 2100%. She said the scale of activity in late September was unprecedented compared to the previous 12 months. Trump called for the technology to be renamed

during his speech to the United Nations General Assembly in New York on September 11th. The use of the word artificial makes it sound fake, he said. It's not fake. It's actually amazing, but we have to be careful. Welcome to the new world of superintelligence. Internet domains are the unique human readable addresses. People type into their web browsers to access websites like BBC.co.uk. Herman said registrations for dot-si web addresses in September stood out very clearly against the other months in the year, although she was cautious about attributing the entire increase to just Trump's comments. She said on September 30th alone, 11,000 new registrations were recorded. Trump hosted a meeting of top technology bosses for what he called an extraordinary

gathering of the people building America's golden age on Tuesday. He signed an executive order instructing U.S. government departments and agencies to start using the term SI and superintelligence and no longer acknowledge the use of the term AI. SI will take the place of AI and official correspondence, websites, reports, and other forms of communication. The order says, said. Images of the signatures for the order were mocked online. However, when it was discovered, United States had been spelt incorrectly under Trump's signature, end quote. Time for the weekend long-range suggestions. U.S. authorities just arrested a California man, suspected of smuggling more than $300 million worth of restricted Nvidia AI chips to China via Malaysia and Singapore.

Bloomberg says that mounting cases of Nvidia chips reaching Chinese companies, despite U.S. export controls, are increasingly pointing to gaps in Nvidia's due diligence. Quote, scrutiny is growing around countries such as Thailand, Malaysia, and Singapore host to legitimate data centers offering remote computing access to Chinese companies, but also transit points for alleged diversion. I've never felt like we got a straight answer as to how it's possible for so many chips to be going to the Southeast Asia region. if they were not serving customers in China, said Daniel Remler, who led AI policy at the State Department under the Biden and Trump administrations. There's just no attempt to grapple with the issue. Bloomberg reported last year that operators in China's western deserts were building dozens of data centers intended to house 115,000 restricted Nvidia processors. At the time, the chipmaker and many government authorities said the plan was unrealistic given Washington's export controls.

but U.S. officials now believe China has smuggled that much computing power and then some, according to people familiar with the details of public prosecutions. Another indictment was unveiled Thursday involving about $300 million worth of Nvidia chips, and there are more probes underway. The exact size of the black market is hard to measure, but AI research firm semi-analysis, whose reports are often cited by industry executives, including Jensen Wong, said supply chain checks and comparisons of production and installation data show hundreds of thousands of AI chips, have slipped through. That's worth billions of dollars and could fill many small data centers or a handful of hyperscale facilities like those Open AI is building in the U.S., end quote. And if you like this sort of thing, Vanity Fair has a huge in-depth profile of Oracle's billionaire owner, Larry Ellison. As a self-taught programmer, he worked his way into a tech scene that included Steve Jobs, Steve Wozniak, and Bill Gates. Ellison's interests were more structural than his peers and frankly boring.

In the 1970s, he was taken by the newish concept of relational databases, a way of organizing information so any part of it could be connected to any other. Every company on Earth would soon have more information than it could keep track of. Governments too. Ellison and his co-founders would sell them the software to make sense of it. The software executive made up for any dullness of mission by becoming a bombastic media figure and maniacally competitive salesman. through the 80s and 90s he publicly jousted with peers, especially Gates. It's one thing I find most contemptible is Bill's lying. This thing about innovating.

It makes me want to puke, Ellison said in a 1998 talk at Harvard. Microsoft has innovated nothing. Over time, though, the rivalry lost its edge. When Ellison bought the Indian Wells tennis tournament in 2009, he would frequently share his owner's box with Gates and Hurd and his wife, Paula. After Hurd had passed and Gates divorced, Paula and Bill became an item. The business press couldn't resist the image of Silicon Valley's swashbuckling technologist who moonlit as a daredevil, and Ellison gladly provided the material. He flew an Italian fighter jet, collected sports cars and speeding tickets, and surfed until a wave almost broke his neck. Ellison's obsession with yacht racing was all-consuming. He took months away from Oracle to chase sailing's biggest trophies, end quote. No bonus content for you this weekend, but it's coming, I swear.