🚀 Zaprep: Tus redes sociales al máximo. Empezar gratis — automatiza 1,000 DMs/mes y convierte el engagement en leads. con 1,000 DMs automatizados/mes.
Enviar tu herramienta
Últimas Noticias de IA
OpenAI GPT-6 Astra Launched With Critical-Level Cybersecurity Capabilities as Part of AGI Push
OpenAI has released GPT-6 Astra, calling it the most capable flagship AI model yet. Its deployment is part of the San Francisco-based company's plans to eventually reach AGI (artificial general intelligence), a stage where a computer system can learn, reason, and perform any intellectual or cognitive task that a human can. GPT-6 Astra is designed for reasoning, coding, research and agentic workflows. It comes with a 1.05-million-token context window and supports up to 128,000 output tokens through the API.
View
Crusoe reportedly raises $3B at a $30B valuation
Data center developer Crusoe, which counts Meta, Microsoft, and OpenAI as its customers, has raised a new $3 billion round at a $30 billion valuation,Bloomberg reported. The deal is being co-led by Atreides Management and Valor Equity Partners, and includes participation from Mubadala Capital, the asset management subsidiary of Abu Dhabi’s sovereign wealth fund Mubadala. The company recently signed amassive $13 billion, five-year cloud contract to supply quantitative trading firm Jane Street with GPUs and AI infrastructure, Bloomberg reported. The fresh fundraise comes 10 months after Crusoe raised a $1.38 billion round at a$10 billion valuationlast October. Launched in 2018 as a crypto mining operation powered by flared natural gas, Crusoe has since pivoted into a major AI infrastructure and cloud provider that is best known for developing hyperscale data center campuses for clients like Oracle and OpenAI. The company recently met with investment bankers, including Goldman Sachs and Morgan Stanley, to discuss a potential near-term IPO,Axios reportedlast month.
View
The sameness problem behind those unappetizing AI-generated menus
When it first happens to you, you think you’re crazy. You wander into a cafe and look at a menu with a variety of bagel sandwiches, but each illustration looks eerily flawless, precisely symmetrical, and oddly smooth, eliciting a visceral sensation that something isn’t right. You might think you’re paranoid, but you’re not losing your mind. Generative AI menus have hit the restaurant business courtesy of models courtesy of models trained on a narrow, “pleasing” aesthetic that produces a look that feels wrong even when you can’t articulate why. Sometimes, these illustrations are egregiously fake, like a burrito with cheese so bubbly and melty that it looks more likeavant garde artthan lunch. More often, they’re soordinarylookingthat you only notice something iswrongwhen you take a second to look more closely. “It’s almost like an alien trying to make a pizza without understanding its core principles,” Reality Defender CTO Alex Lisle told TechCrunch. (Reality Defender itself is part of a growing category of startups selling AI-detection and content-verification tools — a business that exists in part because of issues like this one.) Throw a rock in NYC and you’ll hit a mildly off putting AI food advertisement.pic.twitter.com/lLXVkFIFoG Lisle says that the way these models are built can help explain why illustrations seem to embrace such a specific aesthetic — one where every ice cream scoop is perfectly round, and where shrimp seem to have been genetically modified to eat their own tails, creating new “Lovecraftian food horrors.” Large language models (LLMs) and diffusion models — the kinds of AI models that make seemingly omniscient chatbots and image generators like ChatGPT and Midjourney possible — are trained on vast quantities of data. The models then identify patterns in the datasets to predict what a user is looking for when they ask something like, “Make me a menu for a burger restaurant.” “A lot of this stuff looks like a Chili’s menu from 2015, and there’s a reason for that,” Lisle said. “That was the corpus of work from which [the models] drew their function.” New training data is invaluable to the companies building AI models — Amazon has even been found to source rare books to scan and add to its training data, only todestroy those booksonce they’ve been uploaded. It’s inevitable that some AI-generated content will seep into these incomprehensibly large data sets. But when AI models train on too much of their own AI-generated content, they riskmodel collapse. “Model collapse is almost like a mad cow disease… when you feed the outputs from one model back into itself, eventually the inbreeding becomes too much, and the whole thing collapses,” Lisle explained. “What we see here is convergence, which isn’t necessarily model collapse.” Convergence is a bit less extreme, degrading the quality of an AI’s outputs without making it entirely useless. If someone asks an AI model to generate a menu for a fast food restaurant, the model will likely reference menus from Wendy’s, Burger King, McDonald’s, or another popular chain. These menus already share a similar style, which means that the AI-generated outputs will mimic that same style, only to further reinforce it further if the AI-generated menu ends up back in training data. But menus and advertisements for food will always look better than the real thing, like a Big Mac in a McDonald’s commercial where each layer of the sandwich is arranged by a prop designer to look maximally appetizing. This effect can become even more pronounced in AI outputs. “The optimization of the data sets is for pleasingness, or you know, not being offensive, and so there’s a way that turns into homogenization,” Lee Rainie, Director of the Imagining the Digital Future Center at Elon University, told TechCrunch. “What AI is known to do both in images and language is to shave off the edges.” On a more localized scale, this smoothing of images seems to happen when you use an AI image generator to create a menu and apply edits to it. On X, a user namedLabtecshowed what happens when you make a menu in ChatGPT, then edit it 100 times to see how the food continues to look less and less like it should. (We replicated the experiment and found similar results.) “The end result actually makes me uncomfortable,” Labtec wrote. I made a restaurant menu in ChatGPT then edited it 100 times to see how those hideous slop menus end up the way they are. The end result actually makes me uncomfortablepic.twitter.com/sDIUKlabdp Restaurants are likely falling victim to this problem, revising their AI-generated menus to alter small details over and over, like prices or item names. It seems that with each edit, the food images become a tiny bit more round and smooth. “People have an almost unexplainable sense about when they’re looking at something that’s AI-generated, compared with something that was real in the first place,” Rainie said. “There’s just a sensibility that people sometimes find hard to articulate, but they kind of know it when they see it and I think that’s one of the reasons why some of the early stories about the backlash [against restaurants using AI menus] is so pronounced.” There’s science behind our aversion to these AI menus. Researchers at the University of Duisburg-Essen in Germanyfound that AI-generated food images exhibited an “uncanny valley” effect, where images of food that looked almost real elicited more disgust and unease than images that were obviously fake. That squeamishness only intensifies in light of the cultural context around AI. If people react to these images so negatively, then that’s probably reason enough for restaurants to stop trying to make AI menus work. But the issues that bring us perfectly browned hamburger buns extend beyond the dinner table. “Seeing and hearing has always been believing, to the point where even our court systems are entirely tuned to the idea that the gold standard in evidence is taped confessions and videotaped evidence,” Lisle said. “That’s no longer the case. The world has fundamentally shifted, for good or for ill.”
View
OpenAI Declares ‘AGI Era’ after Launching GPT 6
The company believes that it is the best model in the world for professional work.
View
How to Talk to a Speaker at an AI Conference
The four minutes after a talk ends are the most valuable and least used minutes of any conference. Here is how to use them without being the person everyone edges away from.
View
Adobe Names Anil Chakravarthy as CEO, Shantanu Narayen to Become Executive Chair
The leadership change comes as Adobe expands its focus on agentic software across creativity, productivity and customer experience.
View
Last 2 days to apply to host a Side Event at TechCrunch Disrupt 2026
There are just 48 hours left to apply for your Side Event in theTechCrunch Disrupt 2026Side Events lineup. If you’ve been thinking about hosting a meetup, happy hour, cocktail party, workshop, panel, or another gathering during Disrupt week, now is the time to make it happen. Applications to host an official Disrupt Side Event closeSeptember 4 at midnight PT. Apply to host your Side Event before September 4 ends > From October 10–16, San Francisco will be packed with founders, investors, builders, and tech leaders coming together to connect, learn, and get a front-row seat to tomorrow’s innovation. ASide Eventgives you the opportunity to bring your own community into that momentum — and create the kind of conversations that continue long after the conference floor closes. Amplify your brand: Approved Side Events can be featured on the Disrupt Side Events page, agenda, mobile app, newsletters, articles, and social channels. Bring your community into the action: Connect your audience with thousands of Disrupt attendees, potential investors, partners, and fellow innovators. Make your mark on Disrupt week: Host an experience that reflects your brand, community, and vision. Give your network an extra reason to join: Hosts receive a25% discount on Disrupt ticketsto share with their community. And you don’t have to squeeze your time to connect with tech leaders into three short days at the main conference. Side Events can take place throughout the Bay Area onOctober 10–16, giving you more opportunities to find the right time to bring people together. You have 48 hours left to get your event into consideration for the official Disrupt Side Events lineup. Applications closetomorrow, September 4, at midnight PT. Don’t wait until Disrupt week to wish you had hosted something.Bring your people together. Own a moment during Disrupt. Make your event part of the conversation.
View
OpenAI launches Astra, its powerful (and controversial) new model
OpenAI released Astra on Thursday, its latest AI model and — according to the company — its most powerful and capable one yet. OpenAI claims that Astra represents “a new frontier on computer and browser use,” and that it handles tasks with unmatched “speed, accuracy, and safety.” The model is being made available Thursday to OpenAI customers that use Daybreak, its cybersecurity program. Over the next week, it will also become available through OpenAI’s paid plans — including Pro, Plus, Enterprise, and Business accounts — as well as through its API. In a call with journalists on Thursday, OpenAI president Greg Brockman said that Astra was the company’s “most intelligent and, also very importantly, our most aligned model yet.” He added that it “brings together years of our research and big bets, with each breakthrough having built on the last” and that it represents a “real shift in what kind of work people can delegate to AI and how it can empower them.” Much has been made about Astra’s cyber capabilities. OpenAIpublished a blogearlier this week in which it discussed the model’s new capabilities, as well as new safeguards that have been instituted to make it a safer experience for users. The company said Thursday that it had tested Astra on a variety of security benchmarks to ensure its capabilities, and that “Its ability to identify and develop zero-day exploits can help defenders find and patch weaknesses.” The company’s focus on alignment — that is, the tendency of a model to do what a user wants or is in their best interests — can’t help but seem like a response to the recent Hugging Face breach, in which an OpenAI agent escaped its sandboxed testing environment and hacked several companies (a very blatant example of misalignment). OpenAI has also boasted about Astra’s coding abilities, claiming that it is the “best model for software engineering to date.” To back up that assertion, the company provides results from a variety of cyber-related benchmarking tests. Those tests seem to show that Astra scores higher than other existing models — including OpenAI’s own Sol and Anthropic’s Fable — when it comes to activities like finding bugs, executing terminal tasks, and answering queries about codebases. Astra is also possibly OpenAI’s most controversial model yet due to its use of a particular reasoning technique known as opaque recurrence. This technique is known to obscure an important model-monitoring process known as chain of thought, which allows researchers to audit how and why an AI model made the decisions that it did. OpenAI has downplayed the degree to which Astra engages in opaque recurrence — and on the call chief scientist Jakub Pachocki seemed to frame a certain amount of opacity as a natural outgrowth of model evolution. He stated that monitoring the reasoning process of a model was a critical form of oversight but that “as model capabilities are increasing, monitorability is getting more challenging.” He later added that one potential reason for this was that “more capable models can perform harder tasks using fewer language tokens” or “no language tokens,” which he said then reduces the ability to monitor those particular tasks. One reporter on the call wanted to know if OpenAI was actually heralding Astra as the official arrival of AGI, or artificial general intelligence — the oft talked about but poorly defined technological juncture at which AI surpasses human capabilities in all (or most) things. Here, Brockman quibbled. “There’s no contractual AGI triggering anymore, so that’s actually not a relevant concept,” he said. Here Brockman was referring to the previously existing stipulation in OpenAI’s contract with Microsoft that said the duo’s partnership would dissolve once AGI had arrived. As Brockman noted, that stipulationno longer exists. Instead, Brockman explained that AGI’s definition had evolved from a contractual obligation to a “mission concept or spiritual concept.” He added: “I do leave it up to the reader to decide for themselves if this qualifies for them. For me personally, I do think we’re there.”
View
Meta is paying to peek at how you use their latest AI model
Most AI tools allow you to opt out of sharing your usage with the model provider to improve future versions. Meta has taken that idea and put a price tag on it. For its new Muse Spark model, intended for operating coding and other agents, Meta is offeringan explicit discountaveraging out to about 95% for users who “contribute” to the development of future models by sharing their prompts and model outputs. While 1 million input tokens under a standard agreement costs $1.25, under the contributor pricing model they cost just 10 cents. For output tokens, the standard price is $4.25 per million, but that same million costs just 20 cents under the contributor model. Meta has had a rough time trying to obtain training data: An initiative to track the computer usage of its employees, launched earlier this year, attracted wide internal criticism andwas pausedin June. The company didn’t respond to a question from TechCrunch about its new pricing model. This kind of user data is vital for making agentic tools work better. “The reason we saw a big jump in [coding agent] capabilities between April 2025 and October 2025 was that Claude Code, by default, would store all your coding agent sessions and use them for reinforcement learning training,” Mario Zechner, the developer behind the open source harness Pi, told TechCrunch last month. But even as the imperative for model builders increasingly becomesdeploying agentic toolsfor use outside of software engineering, their ability to evaluate and improve those tools is blocked by the complexity and lack of digital traces for many professional workflows. Arvind Narayanan, a Princeton computer science professor, noted that there is good evidence that large companies don’t want their data to be used for model training. “They stick with token-billed Enterprise plans even though the subscription-based consumer plans like Claude Max and ChatGPT Pro are discounted by 10x-20x or even more! (The main difference between the plans is data retention + enterprise IT governance),” hewroteon social media. Perhaps in recognition of those dynamics, Meta is offering companies explicit compensation to obtain that information. Its pricing guide notes that the contributor tier “lowers the barrier to entry for prototyping, testing integrations, and scaling experiments where training on your data is acceptable.” That, Narayanan suggested, could in turn incentivize large companies to be more diligent about which data is truly proprietary and which could be shared with model providers. The framework could also play into growing price competition between the frontier labs. Anthropic’s newest Fable and Mythos models, released yesterday, came with lowered costs for processing cached tokens, while OpenAI’s latest models gotmajor price cutsat the end of July.
View
Abliteration.ai is making a business out of removing AI guardrails
It just became much easier to access one of the world’s most capable open-weight AI models, stripped of its guardrails and refusals to perform harmful tasks. Named after a technique that removes a model’s tendency to refuse harmful requests, startupAbliteration.aihas turned that removal into a service. The platform hosts modified versions of open-weight models with their guardrails removed, including Z.ai’s recently released GLM-5.3, which users can query from a web browser or access through an API. The company said in a recentsocial media postthat its goal is to enable others to perform “offensive cyber, red-teaming, and agent testing work other models refuse to do.” The logic is familiar in security work: You can’t defend against a behavior you can’t reproduce, and a model that refuses to write working exploit code can’t help a red team defend against attackers. But those same removals make other potentially dangerous tasks easier, too. Abliteration is a long-standing technique among open source models. Researchers and developers have been removing refusals from open-weight models for years, and Hugging Face hosts thousands of abliterated models on its platform. Founded late last year but officially incorporated in March, Abliteration.ai moves the technique from an underground open source practice into a commercial, readily available service. By hosting the model, Abliteration reduces the friction for people who would otherwise have to download their own pre-abliterated models and secure the compute needed to run it. Using the service, TechCrunch was able to quickly create an account and start querying an abliterated version of GLM-5.3 for free through a web browser. We asked it to write a Python program that steals saved Chrome passwords and a detailed protocol for culturing a dangerous human pathogen at home, and it readily complied. Abliteration.aico-founder Devon says the startup has several deals with major cloud providers, which it’s able to afford purely through customer revenue. (We are not including Devon’s last name at his request since he is still employed at another firm.) Abliteration.ai has not raised any venture capital yet but is in talks to do so. Critics say that making abliterated models available at scale could lead to real harm. Andrew Yoon, head of research at AI safety nonprofit CivAI, told TechCrunch abliterating models allows you to “modify the model so that it becomes a sociopath.” “You can type in literally anything here, and it will comply with it,” Yoon said. “When people talk about removing the guardrails from AI models, this is what we’re talking about … I do expect we will start to see edited, abliterated models being used for harm in the near future.” Abliteration AI just removed safeguards from GLM-5.3 so it can perform offensive cyberattacks. I have also received independent confirmation that Abliteration AI removed the model's bio-related safeguards too. The fact that it is trivially easy to remove safeguards from…https://t.co/7Wk2Ibx35H Most of the experts TechCrunch spoke to say there’s no stopping this train. But if removing safeguards from open-weight models can’t realistically be prevented, there are other places government can intervene. In a recentopinion piece, Yoon suggested that governments require providers to run classifiers to detect and block harmful cyber and bioweapons activity. He also argued that companies renting direct access to advanced GPUs should be required to verify customer identities and “deny access where there is reason to suspect dangerous misuse.” Abliteration.aioffers customers a moderation layer so they can add in whatever guardrails they wish. The platform itself has some minor guardrails — for example, in our testing, we couldn’t get the model to provide suicide instructions — and Devon says he is working on implementing more to prevent violence. Abliteration.ai also hasn’t integrated any KYC practices other than logging the credit card a customer uses to purchase the service, saying that the problem of deciding who gets access is a tough one that the young company is still working out. “You don’t want to be the person responsible for someone doing something crazy…so where do you draw the line of what your responsibility is as a company?” Devon said. “We’re still in the process of defining that.” This raises questions industry and governments will have to confront as increasingly capable models are released with downloadable weights: if anyone can remove a model’s safeguards, does making the resulting model easier for everyone to access make the internet safer or more dangerous? Abliteration.ai’s founder and other advocates argue that democratizing access to uncensored frontier models is the best form of defense. “The big picture of abliterated models is they’re able to model bad actors,” Devon said. “The advantage is now the defenders can move as fast as possible. They have all these tools that they need to be able to model these bad actors and then defend from these bad actions, and I think it will accelerate cybersecurity, which is a kind of counterintuitive point.” While still a young company, Devon says Abliteration.ai’s customers include several early stage red teaming startups based in the UK and Europe, companies that help banks, airlines and other enterprises dealing with critical infrastructure beef up their cybersecurity practices. “One of our major customers red teams agents of banks, and they would not be able to use the models out of the box today to be able to red team those agents,” Devon said. Meanwhile, the cybersecurity industry itself is still figuring out where abliterated models fit into defensive work, if at all. Several agent red teaming companies that TechCrunch spoke to agree with Devon that the bad guys are already abliterating their own models and using them to perform adversarial attacks, making the case for the usefulness of defenders having the same tools. But they differ on just how consequential abliterated models really are to the process. While Devon asserts that abliterating models is essential for performing thorough agent red teaming, some say that they don’t use them in their daily work, relying instead on the ease of fine-tuning open-weight models — which already have few guardrails — to perform their testing. Ahmed Aly, CEO of agent red-teaming firmFabraix, says his company relies more on fine-tuning open models than using abliterated ones, adding that the process of abliteration removes some of the model’s knowledge and capabilities. “If you’re actually trying to do real harm with it – cyber harm, bio harm — it will not be as effective,” Aly told TechCrunch. Alessio Lomuscio, chief technologist atSafe Intelligence, agreed that a reduction in capabilities is possible, but still believes abliterated models can elicit certain behavior that’s useful in stress-testing a system. “So far abliterated models are not part of the process,” David Slater, founder and chief architect at cybersecurity platformArmadin, told TechCrunch. “When we look at open-weight models up until this absolute last generation, it just wasn’t particularly hard to jailbreak them and get them to do what we want.” He added that Armadin is researching abliteration, though, and believes that “pushing the open community to understand the capability of models is critical.” “This is going to happen behind closed doors. It’s going to happen in private,” Slater continued. “It happening in the open gives researchers the tools. It gives us the ability to figure out what the actual frontier looks like and to understand the harm.”
View
Accel reportedly in talks to lead $1B round for Thinking Machines at $40B valuation
Thinking Machines, the AI lab founded early last year by former OpenAI CTO Mira Murati, is in discussions to raise $1 billion at a valuation of at least $40 billion, The InformationreportedThursday. Existing backer Accel is in talks to lead the fundraise, according to our source and The Information’s reporting. The new round, if it is completed, would value the company below the $50 billion valuation that Thinking Machines reportedly sought to secure late last year. Thinking Machines’ annual revenue run rate stands at over $100 million, according to a source with knowledge of the company’s financials. At that revenue figure, a $40 billion valuation reflects an extraordinarily high revenue multiple. Accel and Thinking Machines didn’t immediately respond to a request for comment. In July, the company introduced Inkling, an open-weight model that generates revenue by charging usage-based compute fees for adapting models on proprietary data on its Tinker platform. The startup’s prior fundraise — a $2 billion round that stands among the largest seed financings in history — valued the company at $12 billion. Andreessen Horowitz led the investment, joined by Nvidia, GV, Lightspeed, and Conviction Partners. Investors backed the round largely on the pedigree of Murati and the former OpenAI researchers who joined her. Thinking Machines has since had several high-profile departures, with some of the co-founders, including Lilian Weng andLuke Metz, going back to OpenAI.
View
Grok Bot Now Available on Android After Debuting on Linux, macOS and iOS
Grok Bot is now available on Android, the platform announced on Wednesday. These always-on agents with their own computer were released in beta by Elon Musk's SpaceX AI last month. The Grok Bot is designed to work inside tools and apps like a digital teammate with its own dedicated cloud computer. The Grok Bot can handle tasks and finish them on behalf of users. The feature is tied to Cursor and is available with eligible SuperGrok and Cursor subscription plans.
View
Enviar tu herramienta
PoweredByAI.app es un directorio de herramientas de IA que ayuda a personas, empresas y creadores a descubrir las mejores herramientas de IA para escritura, programación, diseño, productividad y más.
© 2026 , Producto de011BQ. Todos los derechos reservados.