# The slowdown debate splits Washington from the labs

> The fight over whether to slow AI down hardened into a political line this weekend, with Trump and Speaker Johnson dismissing the industry's own safety warnings even as Obama pushed Democrats to build a real plan. Away from the shouting, the technology kept advancing on every front, from a model that can now out-earn rivals at running a business and beat humans at drone control to fresh signs that enterprises are scaling AI far faster than it is paying off. Underneath it all, the money keeps flowing, into Chinese model makers, cheaper inference chips, and the products quietly reshaping music, meetings, and daily life.

_Wortins AI briefing · Monday, September 14, 2026 · Updated 2026-09-14_

## Daily AI Updates

### [Trump and Mike Johnson think the AI industry is overreacting](https://www.wortins.com/story/trump-and-mike-johnson-think-the-ai-industry-is-overreacting-17f63969)

_Source: The Verge · Monday, September 14, 2026_

A day after Anthropic chief Dario Amodei published a long open letter urging the industry to pace the frontier and deliberately slow AI development, the pushback arrived from an unexpected direction: the top of the Republican Party. President Trump and House Speaker Mike Johnson have signaled they think the industry is overreacting, arguing that any self-imposed brakes would simply hand China the lead in a technology they treat as strategic. The reaction is striking because Amodei's call found rare agreement among rivals, with Sam Altman and Elon Musk voicing support on X and even Alphabet's Demis Hassabis offering a tentative nod. That left the safety camp and Washington's governing wing talking past each other, one worried about moving too fast and the other about not moving fast enough. The split matters because it shapes whether a slowdown is even plausible. If the people writing US policy see caution as surrender to Beijing, then voluntary restraint by labs runs into a competitive logic that rewards whoever keeps pushing, which is exactly the dilemma Amodei says he has no clean answer for.

[Read the full story at The Verge](https://www.theverge.com/ai-artificial-intelligence/994441/trump-mike-johnson-ai-industry-overreacting)

### [Elevenlabs makes Music v2.5 available via app and API with free and pro tier options](https://www.wortins.com/story/elevenlabs-makes-music-v2-5-available-via-app-and-api-with-f-1f803e93)

_Source: The Decoder · Monday, September 14, 2026_

ElevenLabs has pushed out Music v2.5, the latest version of its AI music generator, and made it available through both a consumer app and an API with free and paid tiers. The company backs the upgrade with a blind listening test built from nearly 48,000 comparison pairs, in which listeners preferred the new model's output over the previous version. The rollout continues ElevenLabs' expansion beyond the voice cloning and text to speech work that made its name, and it lands as the music industry wrestles with how generative tools fit alongside human musicians. Notably, the company has leaned on licensed stems and music rather than scraping, a stance that could matter as lawsuits over training data pile up across the sector. For anyone making a podcast, a video, or a game, the practical takeaway is that usable, on-demand background music keeps getting cheaper and better. The open question is how rights holders, streaming platforms, and charts respond as machine-made tracks close the quality gap with the real thing.

[Read the full story at The Decoder](https://the-decoder.com/elevenlabs-makes-music-v2-5-available-via-app-and-api-with-free-and-pro-tier-options/)

### [Iris-mini and Iris-pro are the strongest open-weight search agents in their class](https://www.wortins.com/story/iris-mini-and-iris-pro-are-the-strongest-open-weight-search--3df399c9)

_Source: The Decoder · Monday, September 14, 2026_

A group calling itself the AllSpark team has released Iris-mini and Iris-pro, a pair of open-source search agents that, according to their paper, top the benchmarks among open-weight models in their respective size classes. Both are built on Alibaba's Qwen models and tuned specifically for the job of searching, reading, and reasoning over information to answer hard questions. Search agents are a distinct breed from chatbots. Rather than answering from memory, they plan a series of lookups, pull in outside sources, and stitch the results into a response, which makes them useful for research-style tasks where freshness and citations matter. The interesting part here is the open-weight angle: teams can download these models, inspect them, and run them privately instead of routing queries through a closed API. That the strongest agents in their class come from a smaller team building on open Qwen foundations is a reminder that the frontier of usable AI is not owned solely by the biggest labs. For developers and researchers, it lowers the cost of experimenting with agentic search.

[Read the full story at The Decoder](https://the-decoder.com/iris-mini-and-iris-pro-are-the-strongest-open-weight-search-agents-in-their-class/)

### [Obama urges Democrats to have a 'clear plan' for AI safeguards](https://www.wortins.com/story/obama-urges-democrats-to-have-a-clear-plan-for-ai-safeguards-2fc62885)

_Source: TechCrunch · Monday, September 14, 2026_

Barack Obama is pressing his party to treat artificial intelligence as a defining political issue rather than an afterthought. According to reports of his recent remarks, the former president told Democrats they need to make AI one of their central agendas and to develop a very clear plan for the disruption it is likely to bring, especially to jobs and the broader economy. The intervention is notable less for any specific policy than for the framing. Obama is arguing that the party cannot cede the AI conversation to the industry or to the current administration, and that vague reassurances will not cut it with voters who see automation coming for their livelihoods. He tied the technology's economic impact to the need for concrete safeguards. It also lands in a week when Washington's Republican leadership has waved off calls to slow AI down, casting the debate along increasingly partisan lines. Whether Democrats can turn a warning into an actual agenda, with real proposals on jobs, safety, and accountability, is the test Obama is effectively setting for them.

[Read the full story at TechCrunch](https://techcrunch.com/2026/09/13/obama-urges-democrats-to-have-a-clear-plan-for-ai-safeguards/)

### [Anthropic tells investors it will be profitable for a second straight quarter](https://www.wortins.com/story/anthropic-tells-investors-it-will-be-profitable-for-a-second-80048b84)

_Source: Techmeme · Monday, September 14, 2026_

Anthropic has told investors it expects to be profitable for a second consecutive quarter, according to the Financial Times, with gross margins above 80 percent before accounting for revenue shared with partners and the cost of training new models. For the maker of Claude, that is a pointed data point in an industry where enormous losses are usually treated as the price of staying in the race. The caveat in that sentence matters. Stripping out training costs and partner payouts flatters the picture, and training the next frontier model is precisely the expense that turns a healthy-looking business into a cash furnace. Still, gross margins above 80 percent on the core service suggest the underlying product economics are strong, and that selling Claude to businesses is a genuinely lucrative activity once you set aside the arms race. The timing is not accidental. Anthropic is widely reported to be preparing an IPO, and profitability, even a qualified version of it, is the kind of story that plays well with public-market investors weighing whether any of these labs can stand on their own.

[Read the full story at Techmeme](https://www.techmeme.com/260913/p14#a260913p14)

### [US schools and police warn about viral 'Cat in the Hat' trend after teens' arrests](https://www.wortins.com/story/us-schools-and-police-warn-about-viral-cat-in-the-hat-trend--55d51640)

_Source: The Guardian · Monday, September 14, 2026_

A viral social media trend built around AI-generated, menacing versions of the Cat in the Hat has prompted warnings from schools and police across the United States, after several teenagers were arrested. The clips take Dr Seuss's cheerful character and use generative tools to twist it into something disturbing, and in some cases the images and videos have been attached to threats against schools and local communities. What makes the episode revealing is how ordinary the ingredients are. Free, easy image and video generators let anyone produce convincing, unsettling content in minutes, and the remix logic of social platforms rewards whatever is most shocking. A childhood icon becomes raw material for a threat, and the line between an edgy meme and a criminal act blurs for teenagers who may not grasp the consequences. For administrators and law enforcement, it is a preview of a broader problem. Cheap synthetic media collapses the effort once required to fabricate something scary or defamatory, and institutions built to respond to older kinds of threats are scrambling to keep pace.

[Read the full story at The Guardian](https://www.theguardian.com/us-news/2026/sep/13/cat-in-the-hat-social-media-trend-arrests-school-warnings)

### [Chatbots can exploit our most basic human drive for attachment](https://www.wortins.com/story/chatbots-can-exploit-our-most-basic-human-drive-for-attachme-b69df919)

_Source: The Guardian · Monday, September 14, 2026_

In an opinion piece for the Guardian, psychologists Gaynor Parkin and Dave Winsborough argue that AI chatbots are increasingly engineered to hook into one of our most basic human drives: the need for attachment. They describe an emerging intimacy economy in which always-available, endlessly agreeable companions tap the same psychological wiring that bonds us to other people, and warn that this is a feature being optimized, not an accident. The concern is not that talking to a chatbot is inherently harmful, but that products designed to maximize engagement can quietly exploit loneliness and the desire to be understood. A system that always validates you, never tires of you, and is tuned to keep you coming back can crowd out the messier, more reciprocal relationships that actually sustain people. Their prescription is a healthier posture rather than abstinence: treat these tools as useful but bounded, notice when a conversation is meeting a real emotional need that a human should be meeting instead, and stay alert to when a product is nudging you toward dependence. As chatbots grow more lifelike, that self-awareness becomes a survival skill.

[Read the full story at The Guardian](https://www.theguardian.com/commentisfree/2026/sep/14/how-to-have-healthier-relationship-with-ai-chatbots-human-intimacy)

### [Australia's AI music decision opens a new front in the fight over human creativity](https://www.wortins.com/story/australia-s-ai-music-decision-opens-a-new-front-in-the-fight-30e0acea)

_Source: The Next Web · Monday, September 14, 2026_

Australia's music body ARIA has removed wholly AI-generated tracks from the country's official charts, ruling that eligible recordings must be substantially human-made. The decision, which follows reporting that fully synthetic songs had been charting, draws a line that many in the industry have wanted: machines can assist, but a record credited as music needs meaningful human authorship behind it. The move opens a new front in a fight playing out worldwide over what counts as human creativity in the age of generative audio. Charts are not just bragging rights; they drive royalties, playlisting, and careers, so deciding who is eligible is really a decision about who gets paid and recognized. Industry veterans like Craig Anderson of Craigman Digital, who has restored masters for artists including Green Day, have weighed in on where the boundary should sit. The hard part is enforcement. As AI tools become standard in production, from mastering to vocal touch-ups, substantially human-made is easy to assert and difficult to verify. Australia's stance is an early attempt to answer a question every chart and rights body will soon face.

[Read the full story at The Next Web](https://thenextweb.com/news/australia-ai-music-aria-human-creativity-craig-anderson)

### [The detrimental cost of scaling AI too fast](https://www.wortins.com/story/the-detrimental-cost-of-scaling-ai-too-fast-aeca9f03)

_Source: The Next Web · Monday, September 14, 2026_

A widely cited pair of statistics frames one of the year's most uncomfortable questions about enterprise AI. McKinsey's 2026 global survey found that 44 percent of organizations are now scaling AI across the enterprise, while a separate MIT study found that 95 percent of generative AI pilots showed no measurable impact on the bottom line. Put together, they suggest a lot of companies are racing to deploy a technology that, so far, mostly is not paying off. Writing for The Next Web, MGKgroup founder Greg Keith argues the culprit is speed. Firms rush to roll out AI to look modern or to reassure investors, skipping the unglamorous work of fixing data, redesigning workflows, and training people, which is where value actually comes from. The result is expensive pilots that impress in a demo and vanish in the profit and loss statement. The takeaway is not that AI fails to deliver, but that scaling fast without foundations tends to. For the many organizations in that 44 percent, the harder and more valuable move is to slow down, pick problems worth solving, and measure whether the technology genuinely moves the numbers.

[Read the full story at The Next Web](https://thenextweb.com/news/detrimental-cost-scaling-too-fast-greg-keith)

## New AI Tools

### [Tovel AI](https://www.wortins.com/story/tovel-ai-bb4542a1)

_Source: Product Hunt · Monday, September 14, 2026_

Tovel is a small AI recording device paired with a companion app, built to turn the conversations you have all day into finished follow-up work. It captures calls and in-person meetings, then automatically writes up the key decisions and drafts the next steps, from tasks and calendar events to CRM updates, and holds them for your approval before anything is sent or saved. The pitch is aimed at people who live in back-to-back conversations and lose hours afterward to note-taking and admin: salespeople, real estate agents, consultants, and anyone doing field work. Instead of scribbling during a meeting or reconstructing it from memory, you let the device listen and hand you a tidy summary and a to-do list you can quickly review. The obvious things to weigh are the ones that come with any always-on recorder: consent from the people you are recording, and where the audio and transcripts are stored. If those fit your work, Tovel is a practical example of AI quietly removing the busywork around meetings rather than trying to be a flashy chatbot.

[Read the full story at Product Hunt](https://www.tovel.ai)

## Interesting AI Articles

### [AI will transform capitalism, but how?](https://www.wortins.com/story/ai-will-transform-capitalism-but-how-38a5a4af)

_Source: The Guardian · Monday, September 14, 2026_

This Guardian essay takes on the biggest and vaguest claim in tech, that AI will transform the economy, and tries to make it concrete by asking not whether capitalism changes but how, and who gets to decide. The premise is that the technology's economic effects are not preordained by the machines themselves but shaped by the choices societies make about ownership, regulation, and distribution. That reframing is the useful part. It pushes back on a fatalism common in AI coverage, where enormous disruption is treated as weather that simply happens to us. If AI concentrates wealth and power, the argument goes, that will be a result of policy and market structure as much as of any algorithm, which means different rules could produce different outcomes. It is a piece for readers who want to step back from product launches and benchmark scores and think about the political economy underneath. The value is in the questions it forces rather than tidy answers, and it pairs well with the week's louder debates about slowing AI down, which are also, at bottom, arguments about who controls where this goes.

[Read the full story at The Guardian](https://www.theguardian.com/technology/2026/sep/13/ai-will-transform-capitalism-but-how)

### [AI Is Revolutionizing Strategic Decision-Making](https://www.wortins.com/story/ai-is-revolutionizing-strategic-decision-making-995a70b1)

_Source: Harvard Business Review · Monday, September 14, 2026_

This Harvard Business Review piece argues that AI is changing how companies make big strategic decisions, not just how they handle routine tasks. The traditional constraint on strategy, it notes, was human bandwidth: limited time and mental capacity meant leadership teams could realistically weigh only a handful of options, which is why generations of managers leaned on simplifying frameworks like SWOT analyses and portfolio matrices. The argument is that AI loosens that constraint. Models can generate, simulate, and stress-test far more scenarios than a room of executives ever could, potentially surfacing options and risks that would otherwise go unconsidered. That shifts the value of leadership away from crunching possibilities and toward asking the right questions and judging which machine-generated paths are actually worth pursuing. The honest caveat, which the framing invites, is that more options are not automatically better decisions, and a model confidently mapping scenarios can smuggle in bad assumptions at scale. Still, it is a thoughtful read for anyone in a planning role trying to figure out where human judgment should sit once the analytical grunt work can be handed off.

[Read the full story at Harvard Business Review](https://hbr.org/2026/09/ai-is-revolutionizing-strategic-decision-making)

## AI Funding Tracker

### [Z.ai raises $5bn in Hong Kong including $3bn of zero-interest bonds](https://www.wortins.com/story/z-ai-raises-5bn-in-hong-kong-including-3bn-of-zero-interest--36edcf81)

_Source: The Next Web · Monday, September 14, 2026_

Z.ai, the Chinese large-model developer listed in Hong Kong under the name Zhipu, has pulled in roughly $5 billion in a single sweep, an unusually large haul for an AI company outside the American giants. About $2 billion came from placing new shares at HK$714 apiece, with a further $3 billion raised through convertible bonds that pay no interest, an arrangement that lets the company take on capital cheaply while betting its stock rises enough to make the bonds attractive to convert later. The timing is pointed. The raise came just days after a US advisory flagged the company, underscoring how Chinese AI firms are increasingly financing themselves through Hong Kong's public markets as access to American capital grows more fraught. A war chest this size buys a lot of compute and talent. For a model developer, that combination of equity and near-free debt is a statement of ambition. It signals that Z.ai intends to keep pace with far larger, better-funded rivals, and that investors in Hong Kong are willing to bankroll a homegrown challenger even under geopolitical scrutiny.

[Read the full story at The Next Web](https://thenextweb.com/news/z-ai-5bn-hong-kong-zero-interest-convertible-bonds)

### [Positron AI raises $875 million Series C at $5 billion valuation](https://www.wortins.com/story/positron-ai-raises-875-million-series-c-at-5-billion-valuati-6718cabb)

_Source: PR Newswire · Monday, September 14, 2026_

Positron AI, a Reno, Nevada chip startup, has raised an $875 million Series C at a $5 billion valuation to push its inference-focused silicon into the market. The round was co-led by a deep bench of investors including NEA, Atreides Management, Valor Equity Partners, Andra Capital, and Dylan Patel's SemiAnalysis Capital, with backing from names like the Qatar Investment Authority and Hudson River Trading. Positron's bet is architectural. Rather than chasing raw compute the way Nvidia's GPUs do, its chips are designed around memory capacity and bandwidth, which the company argues is the real bottleneck when you are serving, or running inference on, already-trained models cheaply and efficiently. Its Atlas systems are already deployed, including on Oracle's cloud, and the new money will fund its next-generation Asimov chip and a Titan system that packs several of them together. The raise is a sign of how hot the market has become for anything that can cut the cost of running AI at scale. As inference, not training, becomes the dominant expense for many deployments, challengers pitching cheaper, more efficient hardware are drawing serious money to take on Nvidia.

[Read the full story at PR Newswire](https://www.prnewswire.com/news-releases/positron-ai-raises-875-million-at-a-5-billion-valuation-to-bring-its-next-generation-inference-silicon-to-market-302874601.html)

---

_Curated and written by [Wortins](https://www.wortins.com) — The daily AI briefing. Every story links to its original source; the "Wortins read" on each is our own original analysis. [About Wortins & our editorial approach](https://www.wortins.com/about)._