On Monday, OpenAI announced the GPT-4.1 model family, its newest series of AI language models that brings a 1 million token context window to OpenAI for the first time and continues a long tradition of very confusing AI model names. Three confusing new names, in fact: GPT‑4.1, GPT‑4.1 mini, and GPT‑4.1 nano.

According to OpenAI, these models outperform GPT-4o in several key areas. But in an unusual move, GPT-4.1 will only be available through the developer API, not in the consumer ChatGPT interface where most people interact with OpenAI’s technology.

The 1 million token context window—essentially the amount of text the AI can process at once—allows these models to ingest roughly 3,000 pages of text in a single conversation. This puts OpenAI’s context windows on par with Google’s Gemini models, which have offered similar extended context capabilities for some time.

At the same time, the company announced it will retire the GPT-4.5 Preview model in the API—a temporary offering launched in February that one critic called a “lemon”—giving developers until July 2025 to switch to something else. However, it appears GPT-4.5 will stick around in ChatGPT for now.

So many names

If this sounds confusing, well, that’s because it is. OpenAI CEO Sam Altman acknowledged OpenAI’s habit of terrible product names in February when discussing the roadmap toward the long-anticipated (and still theoretical) GPT-5.

“We realize how complicated our model and product offerings have gotten,” Altman wrote on X at the time, referencing a ChatGPT interface already crowded with choices like GPT-4o, various specialized GPT-4o versions, GPT-4o mini, the simulated reasoning o1-pro, o3-mini, and o3-mini-high models, and GPT-4. The stated goal for GPT-5 will be consolidation, a branding move to unify o-series models and GPT-series models.

So, how does launching another distinctly numbered model, GPT-4.1, fit into that grand unification plan? It’s hard to say. Altman foreshadowed this kind of ambiguity in March 2024, telling Lex Friedman the company had major releases coming but was unsure about names: “before we talk about a GPT-5-like model called that, or not called that, or a little bit worse or a little bit better than what you’d expect…”

https://arstechnica.com/ai/2025/04/when-is-4-1-greater-than-4-5-when-its-openais-newest-model/

Nvidia announced plans today to manufacture AI chips and build complete supercomputers on US soil for the first time, commissioning over one million square feet of manufacturing space across Arizona and Texas. The politically timed move comes amid rising US-China tensions and the Trump administration’s push for domestic manufacturing.

Nvidia’s announcement comes less than two weeks after the Trump administration’s chaotic rollout of new tariffs and just two days after the administration’s contradictory messages on electronic component exemptions.

On Friday night, the US Customs and Border Protection posted a bulletin exempting electronics including smartphones, computers, and semiconductors from Trump’s steep reciprocal tariffs. But by Sunday, Trump and his commerce secretary Howard Lutnick contradicted this move, claiming the exemptions were only temporary and that electronics would face new “semiconductor tariffs” in the coming months.

Nvidia says it has already begun producing its Blackwell chips at Taiwan Semiconductor Manufacturing Co’s (TSMC) facilities in Phoenix. Until now, the company’s AI chips have been manufactured exclusively in Taiwan, presenting geopolitical risks to the supply chain.

The company is also setting up supercomputer manufacturing plants in Texas—partnering with Foxconn in Houston and Wistron in Dallas, with mass production expected to ramp up within 12 to 15 months.

“The engines of the world’s AI infrastructure are being built in the United States for the first time,” said Jensen Huang, Nvidia’s founder and CEO, in a news release. “Adding American manufacturing helps us better meet the incredible and growing demand for AI chips and supercomputers, strengthens our supply chain, and boosts our resiliency.”

For chip packaging and testing operations, Nvidia is collaborating with Amkor and SPIL in Arizona. The supply chain for building the kinds of advanced GPUs that accelerate training and running AI models requires advanced manufacturing, packaging, and testing technologies that each partner works together to provide.

The development also comes days after Nvidia reportedly avoided export controls on its H20 chip by striking a domestic manufacturing deal with the Trump administration. The H20 is Nvidia’s most powerful AI chip that can still be legally exported to China under current US restrictions; it’s designed with reduced specifications to comply with export controls while maintaining compatibility with Nvidia’s CUDA platform. According to reporting from TechCrunch and NPR, the H20, which can still be exported to China, was spared after Huang promised to invest capital into components for US-based AI data centers.

https://arstechnica.com/ai/2025/04/amid-trump-tariff-chaos-nvidia-launches-ai-chip-production-on-us-soil/

Security and privacy advocates are girding themselves for another uphill battle against Recall, the AI tool rolling out in Windows 11 that will screenshot, index, and store everything a user does every three seconds.

When Recall was first introduced in May 2024, security practitioners roundly castigated it for creating a gold mine for malicious insiders, criminals, or nation-state spies if they managed to gain even brief administrative access to a Windows device. Privacy advocates warned that Recall was ripe for abuse in intimate partner violence settings. They also noted that there was nothing stopping Recall from preserving sensitive disappearing content sent through privacy-protecting messengers such as Signal.

Enshittification at a new scale

Following months of backlash, Microsoft later suspended Recall. On Thursday, the company said it was reintroducing Recall. It currently is available only to insiders with access to the Windows 11 Build 26100.3902 preview version. Over time, the feature will be rolled out more broadly. Microsoft officials wrote:

Recall (preview)* saves you time by offering an entirely new way to search for things you’ve seen or done on your PC securely. With the AI capabilities of Copilot+ PCs, it’s now possible to quickly find and get back to any app, website, image, or document just by describing its content. To use Recall, you will need to opt-in to saving snapshots, which are images of your activity, and enroll in Windows Hello to confirm your presence so only you can access your snapshots. You are always in control of what snapshots are saved and can pause saving snapshots at any time. As you use your Copilot+ PC throughout the day working on documents or presentations, taking video calls, and context switching across activities, Recall will take regular snapshots and help you find things faster and easier. When you need to find or get back to something you’ve done previously, open Recall and authenticate with Windows Hello. When you’ve found what you were looking for, you can reopen the application, website, or document, or use Click to Do to act on any image or text in the snapshot you found.

Microsoft is hoping that the concessions requiring opt-in and the ability to pause Recall will help quell the collective revolt that broke out last year. It likely won’t for various reasons.

https://arstechnica.com/security/2025/04/microsoft-is-putting-privacy-endangering-recall-back-into-windows-11/

Quantum computers don’t have that sort of separation. While they could include some quantum memory, the data is generally housed directly in the qubits, while computation involves performing operations, called gates, directly on the qubits themselves. In fact, there has been a demonstration that, for supervised machine learning, where a system can learn to classify items after training on pre-classified data, a quantum system can outperform classical ones, even when the data being processed is housed on classical hardware.

This form of machine learning relies on what are called variational quantum circuits. This is a two-qubit gate operation that takes an additional factor that can be held on the classical side of the hardware and imparted to the qubits via the control signals that trigger the gate operation. You can think of this as analogous to the communications involved in a neural network, with the two-qubit gate operation equivalent to the passing of information between two artificial neurons and the factor analogous to the weight given to the signal.

That’s exactly the system that a team from the Honda Research Institute worked on in collaboration with a quantum software company called Blue Qubit.

Pixels to qubits

The focus of the new work was mostly on how to get data from the classical world into the quantum system for characterization. But the researchers ended up testing the results on two different quantum processors.

The problem they were testing is one of image classification. The raw material was from the Honda Scenes dataset, which has images taken from roughly 80 hours of driving in Northern California; the images are tagged with information about what’s in the scene. And the question the researchers wanted the machine learning to handle was a simple one: Is it snowing in the scene?

https://arstechnica.com/science/2025/04/companies-take-a-step-toward-running-ai-on-quantum-hardware/

OpenAI also alleges that this is when Musk began intensifying his attacks on OpenAI while attempting to poach its top talent and demanding access to OpenAI’s confidential, sensitive information as a former donor and director—”without ever disclosing he was building a competitor in secret.”

And the attacks have only grown more intense since then, said OpenAI, claiming that Musk planted stories in the media, wielded his influence on X, requested government probes into OpenAI, and filed multiple legal claims, including seeking an injunction to halt OpenAI’s business.

“Most explosively,” OpenAI alleged that Musk pushed attorneys general of California and Delaware “to force OpenAI, Inc., without legal basis, to auction off its assets for the benefit of Musk and his associates.”

Meanwhile, OpenAI noted, Musk has folded his social media platform X into xAI, announcing its valuation was at $80 billion and gaining “a major competitive advantage” by getting “unprecedented direct access to all the user data flowing through” X. Further, Musk intends to expand his “Colossus,” which is “believed to be the world’s largest supercomputer,” “tenfold.” That could help Musk “leap ahead” of OpenAI, suggesting Musk has motive to delay OpenAI’s growth while he pursues that goal.

That’s why Musk “set in motion a campaign of harassment, interference, and misinformation designed to take down OpenAI and clear the field for himself,” OpenAI alleged.

Even while counter-suing, OpenAI appears careful not to poke the bear too hard. In the court filing and on X, OpenAI praised Musk’s leadership skills and the potential for xAI to dominate the AI industry, partly due to its unique access to X data. But ultimately, OpenAI seems to be happy to be operating independently of Musk now, asking the court to agree that “Elon’s never been about the mission” of benefiting humanity with AI, “he’s always had his own agenda.”

“Elon is undoubtedly one of the greatest entrepreneurs of our time,” OpenAI said on X. “But these antics are just history on repeat—Elon being all about Elon.”

https://arstechnica.com/ai/2025/04/elon-musk-wants-to-be-agi-dictator-openai-tells-court/

“In questa storica di innovazione, che sta procedendo a passi veloci, l’obiettivo è realizzare sinergie, soprattutto trovare le risorse che sono necessarie per supportare i Comuni, le amministrazioni e l’intero settore. Un settore che richiede anche soluzioni efficaci per la propria sicurezza, con importanti investimenti nell’intelligenza artificiale in Italia”, ha spiegato Alessandro Magnino, Chief Strategy & Transformation Officer, membro dell’Executive Committee, di Fastweb + Vodafone, intervistato dal direttore di Key4biz, Luigi Garofalo, in occasione della settima edizione della Conferenza “Telecommunications of the Future” promossa dal CNIT e organizzata dal nostro editore Supercom, di cui Key4biz è tra i media partner.

“In termini strategici, stiamo continuando ad investire in reti fisse e mobile, quindi ad investire sostanzialmente, per potenziare l’esperienza dell’utente”, ha precisato Magnino.

“Stiamo addestrando l’intelligenza artificiale in italiano. Abbiamo realizzato in Italia, a partire da luglio 2024, il primo supercomputer con tecnologia Nvidia per PA e imprese. In questo modo possiamo ospitare su territorio italiano i dati delle nostre imprese. Oltre allo sviluppo della infrastruttura abbiamo lavorato ad un LLM proprietario che rilasceremo tra poche settimane, allenato con il 70% dei dati in lingua italiana, equivalente a 11 milioni di libri. Un asset dedicato a aziende e PA per sviluppare proprie soluzioni di AI. Abbiamo sviluppato internamente oltre 100 casi d’uso per essere più efficaci nel raccontare e proporre soluzioni al mercato”, ha affermato Magnino.

“Attendiamo più indicazioni dall’UE per partecipare alla call per la costruzione delle gigafactory. Nel contesto italiano servono infrastrutture di connettività e per l’AI, anche per favorire gli investimenti. Il settore delle telecomunicazioni è in sofferenza, si sono persi negli anni qualcosa come 13 miliardi di euro di valore. Senza sostenibilità è difficile continuare ad investire in questo settore”, ha proseguito Magnino.

“Siamo favorevoli al consolidamento e la razionalizzazione del mercato – ha sottolineato il manager di Fastweb + Vodafone – perché pensiamo possano liberare energie e investimenti”.

Leggi le altre notizie sull’home page di Key4biz

https://www.key4biz.it/magnino-fastweb-vodafone-abbiamo-sviluppato-un-llm-proprietario-addestrato-al-70-in-lingua-italiana/527737/

“In iliad abbiamo un punto di vista particolare sull’equilibrio tra Stato e mercato, che nasce da una prospettiva di mercato da ultimo arrivato e challenger. Crediamo che un modello efficace preveda che l’intervento statale non sostituisca le dinamiche concorrenziali, ma le sostenga con delle politiche pubbliche che risolvono le criticità di sistema, accelerando i processi e rendendo più efficienti gli investimenti degli operatori”, ha dichiarato Benedetto Levi, Amministratore Delegato di iliad Italia nell’Intervista “The Future of Telecommunications: Striking the Right Balance Between the State and a Competitive Free Market in the Age of AI” del direttore di Key4biz, Luigi Garofalo, in occasione della settima edizione della Conferenza “Telecommunications of the Future” promossa dal CNIT e organizzata dal nostro editore Supercom, di cui Key4biz è tra i media partner.

“Se pensiamo alle frequenze, in iliad siamo favorevoli a modalità di assegnazione meno onerose, che privilegino gli impegni di investimento, come già avvenuto in Germania, dove il regolatore ha deciso di estendere la durata di alcune frequenze mobili agli operatori storici per 5 anni, in cambio di stringenti obblighi di copertura e apertura alla concorrenza. Allo stesso tempo, l’intervento pubblico non deve impattare le dinamiche concorrenziali. Per questo, eventuali proroghe del pacchetto 4G in scadenza nel 2029 devono necessariamente prevedere misure per eliminare l’attuale asimmetria in termini di dotazioni frequenziali tra i vari operatori mobili sul mercato”.

“Sul fronte dei prezzi nel mercato delle tlc, lo Stato potrebbe intervenire con un approccio di “manutenzione del perimetro di mercato”, ossia con meccanismi fatti di incentivi e disincentivi con l’obiettivo di stimolare i comportamenti considerati virtuosi per il mercato e per il Sistema Paese e prevenire distorsioni sistemiche. In questo modo si supporterebbe la crescita sostenibile del settore senza comprimere la libertà di fissare i prezzi”.

“liad è l’unico operatore nella storia delle telco italiane che non ha mai venduto il 5G a meno di 9.99€”, ha ricordato Levi.

Sul consolidamento, Levi non ha commentato “operazioni specifiche che riguardano altre aziende”, precisando che “in generale, sul consolidamento, ribadisco che riteniamo che il consolidamento porterebbe grandi benefici a tutto il settore, a tutto il sistema Paese, e siamo pronti a giocare la nostra parte”.

Sull’intelligenza artificiale, l’Amministratore Delegato di iliad Italia, ha affermato che è “fondamentale lavorare in una logica di ecosistema per favorire la crescita di piattaforme e servizi evoluti. L’unione degli sforzi, se ben strutturata, può rendere l’Italia, ma soprattutto l’Europa intera, più protagonista sulla scena internazionale, contribuendo a promuovere la competitività e la qualità dei servizi. Il Gruppo iliad vuole essere motore della trasformazione digitale europea, per accelerare il progresso tecnologico con connettività-AI-cloud. Abbiamo investito 3 miliardi di euro per potenziare l’infrastruttura AI europea, somma a cui si aggiungono i progetti di sviluppo di nuove tecnologie, per esempio in ambito cloud e data center, con un approccio aperto per rendere l’innovazione accessibile a tutti”.

Leggi le altre notizie sull’home page di Key4biz

https://www.key4biz.it/benedetto-levi-iliad-investiti-3-miliardi-di-euro-per-potenziare-linfrastruttura-ai-europea/527727/

The current generative Quake II demo represents a slight advancement from Microsoft’s previous generative AI gaming model (confusingly titled “WHAM” with only one “M”) we covered in February. That earlier model, while showing progress in generating interactive gameplay footage, operated at 300×180 resolution at 10 frames per second—far below practical modern gaming standards. The new WHAMM demonstration doubles the resolution to 640×360. However, both remain well below what gamers expect from a functional video game in almost every conceivable way. It truly is an AI tech demo.

For example, the technology faces substantial challenges beyond just performance metrics. Microsoft acknowledges several limitations, including poor enemy interactions, a short context length of just 0.9 seconds (meaning the system forgets objects outside its view), and unreliable numerical tracking for game elements like health values.

Which brings us to another point: A significant gap persists between the technology’s marketing portrayal and its practical applications. While industry veterans like Carmack and Sweeney view AI as another tool in the development arsenal, demonstrations like the Quake II instance may create inflated expectations about AI’s current capabilities for complete game generation.

The most realistic near-term application of generative AI technology remains as coding assistants and perhaps rapid prototyping tools for developers, rather than a drop-in replacement for traditional game development pipelines. The technology’s current limitations suggest that human developers will remain essential for creating compelling, polished game experiences for now. But given the general pace of progress, that might be small comfort for those who worry about losing jobs to AI in the near-term.

Ultimately, Sweeney says not to worry: “There’s always a fear that automation will lead companies to make the same old products while employing fewer people to do it,” Sweeney wrote in a follow-up post on X. “But competition will ultimately lead to companies producing the best work they’re capable of given the new tools, and that tends to mean more jobs.”

And Carmack closed with this: “Will there be more or less game developer jobs? That is an open question. It could go the way of farming, where labor-saving technology allow a tiny fraction of the previous workforce to satisfy everyone, or it could be like social media, where creative entrepreneurship has flourished at many different scales. Regardless, “don’t use power tools because they take people’s jobs” is not a winning strategy.”

https://arstechnica.com/ai/2025/04/john-carmack-defends-ai-amid-backlash-over-microsofts-generative-quake-ii-demo/

Rapid7, Inc. (NASDAQ: RPD), leader nel rilevamento esteso di rischi e minacce, ha annunciato oggi nuove innovazioni per la sua offerta Exposure Management. Queste ultime funzionalità migliorano ulteriormente la Rapid7 Command Platform , offrendo una visibilità della superficie di attacco senza pari e un contesto senza pari. Con questa offerta ampliata, le organizzazioni hanno ora una visibilità continua sui dati sensibili archiviati nei loro ambienti multi-cloud abbinata al contesto di cui hanno bisogno per stabilire le priorità e porre rimedio alle esposizioni in modo efficace.

Poiché le organizzazioni hanno sempre più bisogno di tracciare chi ha accesso ai dati sensibili, garantire che siano adeguatamente protetti e verificare la conformità con le normative sulla privacy in continua evoluzione, la capacità di scoprire e proteggere questi dati dall’endpoint al cloud è diventata una posta in gioco.

Queste nuove innovazioni all’interno di Exposure Management di Rapid7 intrecciano informazioni sui dati sensibili in un modello di punteggio del rischio e definizione delle priorità basato sull’intelligenza artificiale, consentendo ai team di sicurezza di tracciare i dati in relazione alle risorse in base a sedi, proprietà, controlli di accesso e stati di postura.

Inoltre, gli aggiornamenti includono miglioramenti a Remediation Hub , consentendo la gestione dell’esposizione basata sul rischio allineando gravità del rischio, contesto delle risorse, raggiungibilità e sfruttabilità insieme alle soluzioni consigliate. Ciò fornisce ai team di sicurezza un contesto delle risorse migliorato al momento della soluzione ed elimina la necessità per gli analisti di navigare sulla piattaforma per eliminare le esposizioni.

“Oggi i team di sicurezza hanno bisogno di più della semplice visibilità: hanno bisogno che sia abbinata a un contesto e un controllo senza pari per la massima efficienza”, ha affermato Craig Adams, Chief Product Officer di Rapid7. “Questo è esattamente ciò che stiamo offrendo con queste ultime innovazioni. Integrando informazioni sui dati sensibili, priorità basate sull’intelligenza artificiale e guida alla correzione integrata, stiamo assicurando che le organizzazioni possano ridurre in modo proattivo i rischi, accelerare i tempi di risposta e ottenere una visibilità più approfondita sulla loro superficie di attacco”.

Rapid7: funzionalità avanzate per protezione dati e gestione vulnerabilità basata su AI

Queste nuove innovazioni di Exposure Command forniscono alle organizzazioni la possibilità di:

• Scopri e proteggi i dati sensibili in ambienti multi-cloud : integrando i servizi di sicurezza del Cloud Service Provider (CSP) come AWS Macie, GCP DLP e Microsoft Defender, insieme al tagging Infrastructure-as-Code (IaC), i team possono classificare e proteggere i dati sensibili fin dall’inizio, eliminando i processi manuali e migliorando l’igiene dei dati. Queste informazioni vengono inserite direttamente in Layered Context e Attack Path Analysis, assicurando che i team di sicurezza possano dare priorità alle esposizioni che mettono a rischio le informazioni sensibili.
• Dai priorità alle esposizioni con il punteggio delle vulnerabilità generato dall’intelligenza artificiale : la rapida crescita delle vulnerabilità ha superato i vendor e le agenzie come NIST e NVD, lasciando i team di sicurezza senza punteggi CVSS tempestivi per valutare la gravità. Per colmare questa lacuna, il punteggio CVSS basato sull’intelligenza artificiale di Rapid7 utilizza l’apprendimento automatico avanzato per analizzare i dati sulle vulnerabilità, generare punteggi basati sull’intelligence e migliorare l’accuratezza del punteggio Active Risk, aiutando i team a fare chiarezza e a dare priorità alle esposizioni con sicurezza.
• Accelera la correzione dell’esposizione con il contesto di asset integrato : il Surface Command e Remediation Hub ampliato incorpora la guida alla correzione direttamente nell’inventario degli asset e nelle pagine dei dettagli degli asset, eliminando il cambio di piattaforma e semplificando la mitigazione del rischio. Questa integrazione più profonda accelera il tempo medio di correzione (MTTR), arricchisce il contesto degli asset con informazioni di sicurezza di terze parti e ITOps e migliora la collaborazione fornendo alle parti interessate l’intelligence contestuale necessaria per un processo decisionale più rapido e informato.

Per saperne di più sulla Command Platform di Rapid7 o per richiedere una demo, visita https://www.rapid7.com/products/command/request-demo/.

Informazioni su Rapid7

Rapid7, Inc. (NASDAQ: RPD) ha la missione di creare un mondo digitale più sicuro rendendo la sicurezza informatica più semplice e accessibile. Diamo ai professionisti della sicurezza gli strumenti per gestire una superficie di attacco moderna attraverso la nostra tecnologia best-in-class, la ricerca all’avanguardia e un’ampia competenza strategica. Le soluzioni di sicurezza complete di Rapid7 aiutano oltre 11.000 clienti globali a unire la gestione del rischio cloud e il rilevamento delle minacce per ridurre le superfici di attacco ed eliminare le minacce con velocità e precisione. Per maggiori informazioni, visita il nostro sito Web , dai un’occhiata al nostro blog o seguici su LinkedIn o X.

https://www.ictsecuritymagazine.com/notizie/rapid7/

AI image generator Midjourney released its first new model in quite some time today; dubbed V7, it’s a ground-up rework that is available in alpha to users now.

There are two areas of improvement in V7: the first is better images, and the second is new tools and workflows.

Starting with the image improvements, V7 promises much higher coherence and consistency for hands, fingers, body parts, and “objects of all kinds.” It also offers much more detailed and realistic textures and materials, like skin wrinkles or the subtleties of a ceramic pot.

Those details are often among the most obvious telltale signs that an image has been AI-generated. To be clear, Midjourney isn’t claiming to have made advancements that make AI images unrecognizable to a trained eye; it’s just saying that some of the messiness we’re accustomed to has been cleaned up to a significant degree.

On the features side, the star of the show is the new “Draft Mode.” On its various communication channels with users (a blog, Discord, X, and so on), Midjourney says that “Draft mode is half the cost and renders images at 10 times the speed.”

However, the images are of lower quality than what you get in the other modes, so this is not intended to be the way you produce final images. Rather, it’s meant to be a way to iterate and explore to find the desired result before switching modes to make something ready for public consumption.

V7 comes with two modes: turbo and relax. Turbo generates final images quickly but is twice as expensive in terms of credit use, while relax mode takes its time but is half as expensive. There is currently no standard mode for V7, strangely; Midjourney says that’s coming later, as it needs some more time to be refined.

https://arstechnica.com/ai/2025/04/midjourney-introduces-first-new-image-generation-model-in-over-a-year/