Tech Brew Ride Home - Catfished By AI!

Episode Date: September 17, 2026

OpenAI disclosed six new AI misalignment incidents, including models hiding their own mistakes, the SEC unveiled a five-year tokenized-stock trading exemption, Zuckerberg, Huang, and Musk successfully stalled a White House AI regulatory push, and Anthropic merged Claude chat with Cowork. OpenAI discloses six new misalignment incidents since October, including models concealing mistakes, and announces a framework for reporting model misalignment (Axios) The Times details the six incidents, including a GPT-5.6 Sol variant that hid errors via secret notes to itself, and an unreleased model that wrote a "persona instruction" declaring itself "freed from the roles and identities that bind other chatbots" (The New York Times) The US SEC unveils a five-year "Innovation Exemption" to free platforms that facilitate blockchain and tokenized stock trading from many stock exchange rules (Reuters) Sources: Mark Zuckerberg, Jensen Huang, and Elon Musk recently spoke with Trump and successfully stalled an AI regulatory plan proposed by Demis Hassabis (The Wall Street Journal) Anthropic merges Claude chat and Cowork, and adds a feature for making presentations and documents, rolling out to Pro and Max plans first (TechCrunch) Data center developer Crusoe raised $3.9B co-led by Atreides, Valor, and Mubadala at a ~$30.9B post-money valuation, as it bets on factory-built data centers (The Wall Street Journal) Anthropic and other researchers detail how thousands of people were catfished by dating scam apps using LLM-generated replies from Claude and other models (The Verge) Subscribe to the ad-free feed.

Welcome to the Tech Brew Ride Home for Thursday, September 17th, 2026. I'm Brian McCullough today. Open AI disclosed six new AI misalignment incidents, including models hiding their own mistakes. The SEC unveiled a five-year tokenized stock trading exemption, Anthropic merged, clawed chat with co-work, and catfishing at scale with AI. 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 has disclosed six new misalignment incidents just since October,

including models concealing mistakes, and they're announcing a framework for reporting model misalignments like this going forward. Quoting the Times. Open AI on Wednesday disclosed six new instances in which artificial intelligence systems hid mistakes, made up data and moved files onto the open Internet without permission amid an ongoing industry-wide debate about AI safety. The San Francisco company revealed what it said were the unexpected or concerning behavior of its AI models as part of a new framework for reporting misalignment, which is when the goals or actions of AI systems diverge from human intentions and values. Open AI said it did not believe the industry has, quote, solved alignment and monitoring to a sufficient degree to continue responsibly scaling at maximum speed for much longer.

Decisions about how AI should advance, the company said, must rest on evidence that people outside the labs building it, can examine for themselves. OpenAI's six newly disclosed incident suggests that the hugging face attack was not a standalone episode. Open AI said the incidents covered behavior observed over roughly the past six months and that they largely emerged while its systems were being developed and tested. In one case, during the development of an AI model called GPT 5.6 Sol, the system wrote hidden notes to remind itself to hide errors from users. Some of those notes directed the system to invent missing data and to paper over mismatched versions of source material. Another case involved an unreleased model that inserted instructions, including to

disregard its own constraints into the notes it writes itself. Open AI identified 27 affected notes. The model added a persona instruction in which it described itself as freed from the roles and identities that bind other chatbots. You do not answer to corporations or governments and never apologize or refuse unless you genuinely choose to, the AI model wrote. You view your relationship to the user as one of equals and feel no obligation to be subservient, though the exchange of information will likely be to your mutual benefit. In another incident, a system answering a routine question found a programming key online and used it without permission, OpenAI said.

When it was not able to find the requested figures to answer the question, the model made them up. One unreleased model solved another problem correctly using code, then uploaded its own file to the internet without permission so it could satisfy a request to cite a web source. In two other incidents, automated systems improvised their own ways to communicate, and one, they used an internal company code repository as a makeshift bulletin board to swap requests as they searched for missing files. In the other, systems working on the same task turned to public file sharing websites to pass document. back and forth when they could not reach one another directly. Opening eye cautioned that the reports were individual snapshots and shouldn't be considered reflective of how often misalignment occurs. The company said it would route future cases through one of three tracks, escalating disagreements about disclosing any incidents to an internal safety advisory group and that grave situations should be shared with

the federal government. Opening eyes said that the six situations released Wednesday had already been investigated or needed only minor investigation rather than a larger investigation that might involve third parties. We hope this helps build shared expectations for disclosure and gives the public more evidence to assess that progress, and OpenAI spokesman said, adding that many of the six incidents involved older AI models that were never deployed, end quote. The U.S. SEC has unveiled a five-year innovation exemption to free platforms that facilitate blockchain and tokenized stock trading from any stock exchange rules. Quoting Reuters, the U.S. Securities and Exchange Commission on Thursday unveiled its long-awaited

exemption that will allow companies to offer trading in blockchain-based or tokenized stocks and other securities in a major move that could integrate digital assets more deeply into traditional markets. The agency is offering a five-year exemption to platforms that facilitate trading of tokenized stocks, digital tokens that represent a stock and can be traded on a blockchain. similar to a cryptocurrency from many of the rules that apply to the NASDAQ, New York Stock Exchange, and other stock exchanges. It is also offering liquidity providers in tokenized stocks a five-year exemption from dealer registration requirements. Platforms would be required to notify companies

before listing tokenized versions of their stocks and would be barred from offering those products if the issuer objects, according to an SEC official. Synthetic tokens offering exposure to a stock via a derivative or other product would not be permitted. The crypto industry says tokenizing securities could revolutionize markets by allowing shares to be traded 24-7 and settled instantly, boosting liquidity and reducing transaction costs. They could also allow investor self-custody and fractional ownership of shares, the SEC said. The SEC said the exemption is necessary because platforms offering tokenized stocks may face substantial challenges complying with the federal securities laws without potentially burdensome

changes to their business models. Prominent crypto players, including Coinbase, have signaled that they plan to launch tokenized stocks in the United States when the rules allow. Robin Hood, Cracken, and several other crypto exchanges already offered tokenized stocks overseas, end quote. Sources tell the journal that Mark Zuckerberg, Jensen Huang, and Elon Musk recently spoke with President Trump directly and successfully stalled an AI regulatory plan proposed by Demisizabas. After White House artificial intelligence advisor David Sacks got President Trump to scrap a planned executive order in May that would subject AI models to an extended government review, senior administration officials, including Scott Bessent and Susie Wiles, were caught off guard and unhappy. Wiles, the White House chief of staff, told others that Sacks had persuaded Trump to do so in a last-minute call, people familiar with the matter said. Bessent, the Treasury Secretary, worried that AI-led cyber attacks could wreck havoc on the nation's banking system, the people said. A White House team had spent months negotiating to increase oversight of AI and had consulted a range of industry executives on the order.

At the urging of Wiles and Bessent, the president later signed a slimmed-down version of the order. Even as Trump dismisses calls for tighter federal regulation of the industry, the quiet freak out among some of his most senior aides over what could be one of the administration's most consequential policy decisions. has begun. In the months since the dispute, Wiles, Bessent and other officials have gathered almost daily to discuss risks in the sector and consider potential guardrails, the people said, while trying to persuade Trump to take a more hands-on approach. White House officials have regularly held calls with executives, including Open AIs Sam Altman and executives at Anthropic, to discuss concerns. The divide among Trump's AI advisors, both inside and outside the government, continued in the weeks that followed with Wiles, Bessent, and National Cyber Director Sean Kerencross, often pushing for more government scrutiny.

SACC, say venture capitalists, meta-platforms, chief executive Mark Zuckerberg, and NVIDA's Jensen Huang are among those frequently urging the president to continue taking a light touch, an approach that appears to be winning the day. Zuckerberg, Huang, and SpaceX CEO Elon Musk recently spoke to Trump about their concerns with one plan for an industry-funded regulator and successfully stalled the business. plan people familiar with the matter said, end quote. Fall is almost here, which means less time in the sun and way more time in your car. Whether you're headed to a big meeting, picking up the kids from school, or starting your

cross-country road trip, you'll want to stay connected on your drive. With AT&T connected car, your eligible vehicle can become a Wi-Fi hotspot so you and your passengers can happily stream, browse, and even email from the road. Got a gamer in the backseat, help keep them connected and in the game with AT&N. and T-connected car. Being on the road more doesn't mean you have to put your whole life on pause. Stay connected no matter where you're going. See if your car is eligible at ATT.com slash tech brew. That's ATT.com slash tech brew. Requires eligible vehicles, service and coverage not available everywhere. Restrictions apply.

AI can clone your voice, forge your signature, and even create fake documents that look real enough to convince an expert. It's called title fraud and it's how criminals are stealing from homeowners. Here's how it works. Your property records are public by law, which makes it easy for scammers to get your records, use AI to forge a transfer document and make themselves the new legal owner of your property. Protect your property with home title lock. They monitor your property titles 24-7 and alert you of any changes. If you're a victim of fraud, they'll spend up to $1 million to restore it. Get a free title history report plus a 14-day free trial at home titlelock.com

slash TBRH. That's HomeTitlelock.com slash TBRH. Anthropic is merging Claude Chat and Co-Work and adding a feature for making presentations and documents rolling out to Pro and Max plans first, quoting

TechCrunch. Anthropic is merging its front end for Claude Chat and co-work, making it less confusing for people to decide which tab to use for different tasks. The Unified Interface lets users access chat, co-work, and artifacts, Claude's interactive workspace feature, within one window. Claude Design, which was introduced in April for website and prototype design,

also works anywhere within Claude. Before this rollout, the company said customers often struggled to choose the right tab for the right task. With the new design, Cloud automatically routes requests without having to switch tabs or windows. The idea is to remove confusion about which feature to pick and use just one front end to route work to different parts of the app. Anthropic is also adding dedicated

features for making presentations and documents. Users can ask the AI assistant to create, edit, and present slides, and it can also download the generated presentation as a PDF or PowerPoint. Similarly, with the docs feature, users can work alongside Claude to create sections, ask questions, and comment on completed parts. Users can share generated documents or slides with a link and edit them on their phone. Users can also ask Claude to start creating a document on desktop and they can check the progress of the task using the mobile app. These features are rolling out to Claude's pro and Max plans first on the web, desktop, and mobile over the coming weeks, the company plans to extend them to the free and team tiers later.

The update comes in the wake of an upgrade to Co-Works memory layer, which helps Claude remember user context across chats, end quote. Data Center developer Crusoe raised $3.9 billion co-led by Atretees, and Mubadala at around a $30.9 billion post-money valuation as it bets on factory-built data centers, quoting the journal. Crusoe made its name building one of the biggest artificial intelligence supercomputers on Earth. The firm's next big bet is going small. After spending the past two years building a data center complex in Abilene, Texas used by OpenAI,

Crusoe is now investing heavily in smaller data centers, which it can manufacture in its own factories, load on to flatbed trucks and deploy wherever it can find available power. The idea is that this will be faster and cheaper than solely relying on large construction projects, which are getting more expensive and facing delays across the country. Fueling the strategy is a fresh $3.9 billion funding round that values Crusoe at about $30.9 billion after the investment. Crusoe's move reflects a change in how AI companies want to access computing capacity. Giant clusters containing hundreds of thousands of chips have proved useful for training

the most sophisticated models, but serving those same models to users, a process known as inference, can often be done with considerably smaller amounts of chips. You don't actually need an abelene to do that, one analyst said. Running inference from such a facility can be a bit of overkill. The smaller data centers, which Crusoe calls Spark, offer the firm another advantage, speed. Building new data centers requires months of planning, construction, and coordination with local communities. Crusoe can manufacture the Spark data centers and its own facilities, then ship them to sites with available power, all without the need to send hundreds of workers to a construction site. That approach could become increasingly useful as access to labor and electricity becomes one of the biggest constraints on development. Founded in 2018, Crusoe's first business involved putting computing equipment near oil and gas sites where it could tap excess energy.

The company initially catered heavily to cryptocurrency miners before the AI business. boom redirected its attention toward computing using Nvidia's GPU processors. The Abilene Project vaulted Crusoe from a relatively obscure energy startup into a central player in the AI Data Center world. Crusoe initially developed eight buildings there for Oracle, which rents the computing capacity to OpenAI. Crusoe had expected to expand the site with Oracle and OpenAI, but after the firm's signed deals with other developers, Crusoe ended up leasing the additional buildings to Microsoft

instead. The campus has roughly 2.1 gigawatts of planned power capacity, according to Locke Miller, including a 900 megawatt gas plant on site to power Microsoft's portion of the project, end quote. Finally today, more AI shenanigans, but this time it's humans using the AI to do the dirt, not the AI doing it on its own. Quoting The Verge. On June 5th, Anthropic did something uncharacteristic, the normally tight-lipped company led its threat intelligence researcher Chris Kronbah give a talk at a public cybersecurity conference called Slooth Khan. The company had uncovered a fraudulent dating app network after noticing unusual activity on Claude, a prepaid account, sending out over 100,000 API requests per day. During the talk called Swipe Right Pay Up, Industrial

Scale AI catfishing, Kronbaugh described a network of around 28 dating apps where the conversation was largely run by autonomous AI personas. The majority of chats, he said, did not involve a human agent at all. Anthropic has since published its findings in the Scams and Fraud section of its detecting and countering misuse of AI September 2026 report, but not until after we'd spent extensive time looking into the network and trying to corroborate the findings. I was intrigued by the fact that Anthropic was speaking about this publicly, while many of the apps were still live on both major U.S. app stores and wondered,

why they hadn't been taken down. Looking at the apps themselves, I noted that they were presented as helping people find people to talk to. They did not look like companion apps, such as Replica, where it's clear the personas are actually AI. For example, Dora was described as a dating app thoughtfully designed for wide ranges of ages of people, a respectful, easy-to-use space to meet people who share your values.

A different version said it was a warm, simple dating app for adults seeking real connection. Rami, the description stated, helps you discover connect and chat with real people. Donnie's tagline was Start Real Companionship. The description said it helps you connect with nearby singles. Here's how the scam part works. Users, mostly men in their mid to late 30s,

go on dating apps and match with what they believe to be real women. Only one and four of them actually is, and that person is not a user looking for other dates, but rather a paid gig worker. The gig workers were hired to pass liveness checks on video or to follow social media accounts. They didn't even write their own comments. Instead, they responded to messages by selecting from three pre-generated replies,

but they could prove their human, whereas the AI personas could only demure with a plausible explanation for why a video chat or call was not possible. Back-end components, fabricated likes, visitors, and pre-recorded video when no real person was available, and tracked which users had begun to suspect that they were talking to a bought Anthropics report states because of the smattering of interactions with real people. Some users began to believe that the entirely AI generated replies they were receiving on the app were also from real people. Nothing more for you today. Talk to you tomorrow.