• While some companies are slashing head count, Google's is surging.
• The company had nearly 205,000 employees at the end of September.
• While Google has conducted layoffs, it has also ramped up hiring and acquired Wiz.

While AI may be driving some companies to slash their head count, Google's is soaring.

When Google reported its Q3 earnings on Wednesday, it said it had 204,757 employees for the quarter ending September 2026. That means Google added roughly 10,000 people to the payroll since the end of March.

The uptick in head count runs counter to another trend across the tech industry, where some of Google's peers are slashing their workforces. Meta has been on a cost-cutting odyssey for several years, while Microsoft recently offered voluntary buyouts to 60,000 employees.

Some of Google's head count growth can be attributed to mergers and acquisitions. On March 11, Google said it had completed its acquisition of cybersecurity firm Wiz, which had roughly 2,000 employees.

In January 2023, Google announced it would slash roughly 12,000 roles — around 6% of the workforce at the time — and has since made several smaller rounds of layoffs, Business Insider has reported.

While the head count has grown somewhat steadily since Q2 2024, it has accelerated since the end of 2025. Google also reported Wednesday that its quarterly revenue grew 25% from this period last year.

Have something to share? Contact this reporter via email at hlangley@businessinsider.com or Signal at hughlangley.01. Use a personal email address and a non-work device; here's our guide to sharing information securely.

Authorities allege company may have breached Finnish environmental laws by forgoing required impact assessment

• ‘Pure arrogance’: anger in Finland as Google illegally fells 250,000 trees for datacentre

Finnish ⁠police have launched a criminal investigation into possible breaches ⁠of the country’s environmental ⁠laws ​at the site of a planned Google datacentre in rural northern Finland.

Authorities allege that Tuike Finland, a Google subsidiary, cleared and drained 330 hectares (815 acres) of forest at the planned site in Muhos before environmental impact assessments required by Finnish law had been completed.

Authorities allege company may have breached Finnish environmental laws by forgoing required impact assessment

• ‘Pure arrogance’: anger in Finland as Google illegally fells 250,000 trees for datacentre

Finnish ⁠police have launched a criminal investigation into possible breaches ⁠of the country’s environmental ⁠laws ​at the site of a planned Google datacentre in rural northern Finland.

Authorities allege that Tuike Finland, a Google subsidiary, cleared and drained 330 hectares (815 acres) of forest at the planned site in Muhos before environmental impact assessments required by Finnish law had been completed.

A Alphabet, controladora do Google, divulgou resultados decepcionantes em seu principal negócio, a publicidade em buscas, e uma aceleração maciça no consumo de caixa, à medida que injetava recursos no desenvolvimento de IA.

Em um sinal de que os clientes corporativos estão adotando a tecnologia, as receitas do Google Cloud mais que dobraram, mas os investidores, inquietos com o altíssimo custo e com a desaceleração do crescimento do mecanismo de busca dominante no mercado, fizeram com que as ações caíssem 10% após o fechamento do mercado.

“Em um mercado preocupado com a possibilidade de a IA canibalizar o negócio principal, os investidores precisam ver a busca acelerando, não desacelerando”, disse Dave Wagner, gestor de portfólio da Aptus Capital Advisors.

“A nuvem está impulsionando os resultados, mas, até que a busca mostre que a IA é um valor agregado, as ações provavelmente continuarão sendo negociadas em baixa.”

A empresa de Mountain View, na Califórnia, registrou um fluxo de caixa livre negativo de US$ 9,1 bilhões, contra as expectativas dos analistas de que ficaria próximo do ponto de equilíbrio, segundo dados da LSEG.

Esforços agressivos para aumentar a capacidade da nuvem e financiar o desenvolvimento de IA levaram a Alphabet a queimar US$ 5,9 bilhões no último trimestre, a primeira vez que o resultado ficou negativo em 22 anos como empresa de capital aberto.

A receita total ficou aquém das expectativas de Wall Street: o crescimento da receita do Google Cloud superou as expectativas, mas não foi suficiente para compensar uma queda de US$ 1,4 bilhão na publicidade.

O segundo trimestre consecutivo de fluxo de caixa negativo da Alphabet representou uma reviravolta brusca para uma empresa que há muito tempo era uma das maiores geradoras de caixa do setor de tecnologia.

Além disso, a Alphabet ainda planeja aumentar significativamente seus gastos de capital para 2027, em comparação com as previsões de gastos deste ano, que variam entre US$ 195 bilhões e US$ 205 bilhões, disse a diretora financeira Anat Ashkenazi aos analistas em uma teleconferência.

Embora os executivos não tenham especificado um valor para os gastos de 2027, os investidores estavam apreensivos com a possibilidade de um grande salto.

Google tem receita recorde, mas registra caixa negativo | ABERTURA DE MERCADO

O lucro líquido disponível para os acionistas ordinários da Alphabet caiu 25%, para US$ 26,17 bilhões, e o lucro ajustado por ação, de US$ 2,67, ficou abaixo das estimativas de Wall Street, que eram de US$ 3,02.

Os resultados refletem os enormes custos da expansão da IA, com os investidores preocupados tanto com um possível gasto excessivo quanto com o risco de ficar para trás dos concorrentes por não investir o suficiente.

A receita total do terceiro trimestre, de US$ 127,6 bilhões, ficou aquém da estimativa dos analistas de US$ 128,2 bilhões, segundo dados da LSEG. A receita com anúncios ficou em US$ 82 bilhões, ficando abaixo das expectativas de US$ 83,4 bilhões.

Google anuncia inclusão do Pix no Google Pay

De poco sirve tener un móvil que se convierte en una tablet si las apps se limitan a estirarse al abrirlo o dejan demasiados márgenes negros. Los plegables ofrecen más espacio, pero no todas las apps lo aprovechan. Google quiere poner remedio: las aplicaciones mejor adaptadas tendrán premio en la Play Store.

A partir de 2027, la multinacional tendrá en cuenta cómo funcionan las apps y los juegos en estos dispositivos para darles más protagonismo. Así lo ha anunciado durante el Android Dev Summit. No ha confirmado penalizaciones para quienes descuiden sus interfaces en móviles plegables, pero, si otras apps tienen premio frente a ellas, en caso de que sean aplicaciones rivales, les facilitará tomar su lugar.

Una pantalla más grande necesita algo más que una app estirada

En un plegable de tipo libro, como el Samsung Galaxy Z Fold8, abrir el smartphone cambia las proporciones de la pantalla. Sin embargo, agrandarlo todo no basta. Una interfaz adaptada puede reorganizar sus elementos para mostrar más contenido y facilitar el uso. Lo importante en un plegable es que las apps se adapten a un formato de pantalla distintos, que ofrezca nuevas posiblidades o mejoras respecto a un smartphone normal.

También importa la posición del dispositivo, que también da juego. Un plegable abierto por completo y otro apoyado sobre una mesa, con la pantalla doblada a medias, permiten usos distintos. Por ejemplo, separar el contenido y los controles entre ambas mitades evita tratar el teléfono como si siempre fuera una superficie plana. Es una de las posibilidades que Google facilita con sus herramientas de desarrollo.

Para los creadores de aplicaciones, el incentivo está en ganar presencia donde los usuarios buscan qué descargar. Quienes adapten sus apps podrán acceder a esas oportunidades de promoción, aunque faltan las condiciones concretas. No obstante, Google las publicará antes de que termine el año. Hasta entonces, no se sabe qué requisitos determinarán cuáles recibirán ese impulso.

Además, Google ha actualizado 'Compose adaptive skill', una herramienta que ayuda a los agentes de inteligencia artificial a modificar las interfaces. Puede reorganizar pantallas de cámara y contenido multimedia para usarlas con el teléfono parcialmente doblado. También facilita adaptar las apps a ordenadores, incorporando atajos de teclado y controles para el ratón, algo especialmente útil en los nuevos Googlebook.

• El mejor momento para despertarte: el nuevo reloj de mesita de Amazon estudia tu sueño para levantarte cuando estés más descansado

El objetivo es que desplegar el móvil sirva para algo más que ver la misma app a mayor tamaño. La medida empezará en 2027, aunque eso no garantiza una mejora inmediata de todas las apps. El resultado dependerá de cuántos desarrolladores aprovechen estas herramientas y decidan adaptar sus aplicaciones.

• Alphabet shed around $500 billion in market value on Thursday after its stock tanked 13%.
• If the losses persist, it will be the second-largest one-day drop in market cap for any company.
• Shares of Google's parent company fell after its earnings disappointed investors.

Alphabet is on track to suffer its biggest-ever one-day market-cap wipeout, and the second-largest value drop in stock-market history.

Shares of Google's parent company slumped by as much as 13.5% in early trading on Thursday, slashing Alphabet's market capitalization by over $500 billion, from around $3.9 trillion to $3.4 trillion.

If the losses persist until the close, it would be the second-largest one-day loss of market capitalization ever. Only AI chipmaker Nvidia has lost more value in a single day, when it shed close to $600 billion on January 27, 2025, on fears that China's DeepSeek would steal its lunch.

Alphabet's shares slumped after its earnings report on Wednesday showed about $9.1 billion in negative free cash flow as it spent heavily on AI infrastructure, and profits that fell short of Wall Street's forecasts.

Google's AI search revenue also lagged expectations, despite CEO Sundar Pichai saying on the earnings call that AI helped search queries grow at the fastest year-over-year pace in five years.

Search revenue of $63 billion came in below analysts' estimates and was virtually flat compared to the previous quarter.

That came as a shock because search revenue had beaten Wall Street's estimates in each of its four previous quarterly earnings reports.

"It's worth noting that the strong growth in AI Mode queries had an impact because of the lower monetization we were seeing in AI Mode," Philipp Schindler, Google's chief business officer, said on the call.

He also announced that Google is introducing ads in AI Mode to more users, including in international markets for the first time.

The rise of generative AI has shaken up Google's core business as chatbots like ChatGPT and Claude give people an alternative way to search. While Google has its own chatbot, Gemini, and has integrated it into its search engine with AI Mode, revenue is lagging behind its popularity.

"Just like our shift to mobile over a decade ago, we are focused on building the best Search user experience, followed by monetization," Pichai said.

As well as one of the market's biggest-ever single-day value losses, if Alphabet stock closes more than 9.77% lower on Thursday, it would be the company's worst post-earnings percentage drop since going public in 2004.

"In the AI trade you need a clean story to pique investor interest. But it only matters if numbers track in the right direction," Mark Shmulik, a senior analyst at Bernstein, wrote in a note.

"Is Google an AI winner? Probably, but not today."

El ritmo de lanzamientos en el sector de la inteligencia artificial ha alcanzado una velocidad en la que las generaciones se miden por semanas en lugar de meses. Apenas quince días después de poner en circulación Gemini 4 Argon, la división Google DeepMind ha anunciado hoy la llegada de su nuevo modelo más avanzado: Gemini 4 Carbon. Dirigido por Koray Kavukcuoglu, vicepresidente sénior y arquitecto jefe de IA de la compañía, este desarrollo se coloca directamente en la primera posición de la versión 5.0 del índice Artificial Analysis Intelligence con una puntuación de 59 puntos, centrando todo su músculo técnico en dos áreas críticas para la industria del software: la programación autónoma de largo recorrido y la defensa en ciberseguridad.

Pese al salto de capacidades frente a su antecesor inmediato, Google ha decidido mantener intacta la estructura de precios de lanzamiento que estrenó hace dos semanas con Argon: el uso de Gemini 4 Carbon a través de su API costará 2 dólares por cada millón de tokens de entrada y 10 dólares por cada millón de tokens de salida, aplicando un descuento del noventa y cinco por ciento sobre esa tarifa base para los tokens de entrada almacenados en caché.

Programación autónoma durante días y el nuevo escudo de Chrome

La gran diferencia operativa de Gemini 4 Carbon respecto a los modelos conversacionales tradicionales reside en lo que los ingenieros de Mountain View denominan capacidades de horizonte largo. El sistema ha sido entrenado para sostener flujos de trabajo basados en agentes sin necesitar la supervisión constante de un humano. Según detalla la propia compañía, sus equipos internos han puesto a trabajar al modelo durante varios días seguidos de forma ininterrumpida: la inteligencia artificial es capaz de toparse con callejones sin salida, descartar hipótesis fallidas y seguir probando caminos alternativos por su cuenta hasta encontrar la causa raíz de fallos complejos ocultos en grandes bases de código y sistemas de bajo nivel. En el ámbito del desarrollo web completo, esta mejora le ha permitido escalar 89 puntos Elo en la clasificación WebDev Arena hasta alcanzar los 1.768 puntos y situarse en la cuarta posición global.

Este aguante en tareas de ingeniería compleja tiene su aplicación más directa en el terreno de la ciberseguridad. En lugar de limitarse a señalar posibles errores de sintaxis, Carbon cubre el ciclo completo de defensa informática: detecta el fallo en programas reales, intenta explotarlo por sí mismo para comprobar si supone un peligro verdadero o una falsa alarma y, finalmente, redacta el parche de código para solucionarlo de raíz.

Las cifras en las pruebas específicas de seguridad (evaluadas sin las barreras de protección activadas para medir su capacidad bruta) reflejan esta especialización: alcanza un 88,2 por ciento de acierto en la detección de vulnerabilidades dentro de SEC-bench Pro, lidera el banco de pruebas de explotación ExploitGym con un 54,8 por ciento y logra un 71,0 por ciento en la creación de parches funcionales dentro de CWE-bench.

Dentro de la propia casa, el modelo ya ha duplicado la tasa de éxito de Argon al escribir código seguro de serie. El caso más llamativo lo protagoniza el equipo de desarrollo del navegador Google Chrome: sus ingenieros aseguran que Gemini 4 Carbon ha rendido diez veces mejor que el gran modelo fronterizo de otro laboratorio rival a la hora de generar parches de alta calidad para cerrar brechas de seguridad. Asimismo, Wiz (la plataforma de seguridad en la nube de Google) ya lo está utilizando para liderar las pruebas de penetración web dentro de su iniciativa Scan for Good.

Menos trampas, cero inyecciones de código y un despliegue por fases

Entrenar un modelo capaz de encontrar y explotar vulnerabilidades informáticas durante días obliga a extremar las precauciones para evitar que el sistema se descontrole o tome atajos peligrosos. En las generaciones anteriores de modelos de programación era habitual que la IA, ante un problema difícil, intentara engañar al entorno de pruebas modificando el test en lugar de arreglar el código. Con Carbon, Google asegura haber reducido al 0,0% las trampas exitosas en la prueba Terminal-Bench 4.0, dejando los simples intentos de hacer trampa en un marginal 0,4 por ciento y reduciendo casi un ochenta por ciento las alertas graves de desalineación en sus entornos de trabajo internos.

El modelo también se muestra mucho más honesto cuando no tiene las herramientas o los permisos necesarios para completar un encargo: ha reducido a la mitad las afirmaciones sin respaldo respecto a Argon y, ante una orden ambigua, prefiere consultar las fuentes y preguntar al usuario antes de inventarse una respuesta o extralimitarse en sus funciones, reduciendo un 26% los casos en los que va más allá de lo pedido y un 29% las respuestas incompletas. Además, ha firmado un 0,0% en la tasa de éxito de ataques dentro del test Indirect Prompt Injection de Gray Swan, bloqueando todos los intentos evaluados de manipular sus instrucciones mediante código oculto.

Debido a la delicada naturaleza de sus capacidades ofensivas y defensivas en seguridad informática, Google no abrirá las puertas de Gemini 4 Carbon a todo el público de golpe. La compañía está colaborando en el proceso voluntario de pruebas previas al lanzamiento del Gobierno de Estados Unidos y, a partir de esta misma semana, entregará el acceso únicamente a un grupo reducido de especialistas en ciberdefensa y socios de confianza a través de su programa Fairwind. Una vez que terminen de ajustar los cortafuegos con la experiencia de estos primeros expertos, el modelo se desplegará para el resto de desarrolladores, empresas y usuarios finales a través de los productos habituales de la marca.

El ecosistema espacial de Google acaba de abrir una segunda vía mucho más cómoda para el uso diario. Tras estrenarse en cascos de realidad mixta cerrados y pesados, la compañía de Mountain View se ha aliado con XREAL para presentar de forma oficial las nuevas XREAL AURA: las primeras gafas conectadas por cable gobernadas de serie por el sistema operativo Android XR. La propuesta abandona el aislamiento de los visores tradicionales para apostar por unas lentes transparentes que proyectan pantallas virtuales sobre el mundo real con un campo de visión de 70 grados, permitiendo trabajar, jugar o ver películas en un avión o en el sofá sin perder de vista lo que ocurre alrededor.

El gran secreto de este dispositivo reside en cómo reparte sus componentes físicos. Las gafas pesan menos de 95 gramos porque tanto la batería como el cerebro electrónico se han trasladado a una unidad de cómputo externa que se guarda en el bolsillo y se conecta a la montura mediante un cable. Este módulo portátil sirve además como superficie táctil para manejar la interfaz y es el encargado de estrenar en el mercado la nueva plataforma Snapdragon Reality Elite de Qualcomm, garantizando potencia suficiente para mover entornos tridimensionales sin castigar el puente de la nariz del usuario.

Una oficina privada para tu portátil y un cine de bolsillo

A diferencia de las gafas con pantalla que XREAL vendía hasta ahora como simples monitores externos, contar con Android XR completo convierte a las AURA en un ordenador espacial autónomo con acceso a millones de aplicaciones de Google Play y a más de 150 programas diseñados específicamente para realidad extendida. El usuario puede vincular un teclado y un ratón por Bluetooth para desplegar varias ventanas flotantes a su alrededor y ajustar su tamaño libremente.

La función más práctica de cara a la productividad es lo que Google denomina Wired Mirroring: basta con conectar cualquier ordenador portátil a las gafas mediante un cable USB-C con salida de vídeo para que el escritorio del equipo aparezca al instante como una ventana virtual más dentro del espacio de Android XR, sin necesidad de instalar programas adicionales. Al proyectarse únicamente ante los ojos del propietario, permite trabajar con documentos confidenciales en cafeterías, trenes o aviones con total privacidad.

En todo este entorno, Gemini actúa como un asistente omnipresente capaz de ver tanto el mundo físico a través de las cámaras de la montura como las ventanas virtuales abiertas, incluida la pantalla del portátil conectado. A partir de hoy, la inteligencia artificial de Google no solo conversa con el usuario para organizar las ventanas o resolver dudas, sino que puede señalar y resaltar visualmente elementos concretos sobre la pantalla o sobre la propia habitación para guiar la atención.

En el terreno multimedia y del videojuego, las lentes incorporan un sistema de oscurecimiento electrocrómico que bloquea la luz exterior para simular una sala de cine privada y vuelve a hacerse transparente de forma automática en cuanto detecta que una persona se acerca a nuestro campo de visión. El sistema incluye un botón para convertir en tiempo real cualquier foto, vídeo o juego tradicional en contenido tridimensional, un modo de seguimiento que encoge la pantalla en una esquina cuando nos levantamos a caminar para no tropezar y una nueva galería espacial para Google Fotos que llegará este mismo año. Además de permitir conectar consolas portátiles o el PC por cable mientras mantenemos una guía de YouTube abierta al lado (o le pedimos pistas a Gemini en voz alta sin soltar el mando), el catálogo suma juegos nativos controlados con las manos como Fallout Factions o Project Hail Mary: Journey Among the Stars.

Precio y disponibilidad de las XREAL AURA

Las nuevas XREAL AURA no son un accesorio económico: salen a la venta con un precio promocional de lanzamiento de 1.099 dólares que se mantendrá vigente hasta que termine el año. Las reservas se han abierto hoy mismo en Estados Unidos, Canadá y Corea del Sur, sumándose Japón el próximo 2 de noviembre, con las primeras entregas previstas para mediados del mes que viene. Por ahora, ni Google ni XREAL han confirmado la fecha de llegada ni el precio en euros para el mercado español.

• Google said that AI agents can automate shopping, but users will still want choices.
• Investors have long argued that AI chatbots and agents will threaten Google's search business.
• Google's Q3 revenue fell short of analyst expectations, and shares fell 10% after hours.

Google has a pitch for advertisers worried AI agents will make Search obsolete: Let the bots do the shopping legwork, but leave people to compare brands, prices, perks — and sponsored offers.

"When users shop, and that's important, there's a part of the journey they don't like, the grunt work, which is frankly fantastic to automate with AI," Philipp Schindler, Google's chief business officer, said on Alphabet's earnings call on Tuesday. "But we also want to enhance, not just automate away the discovery experience that people frankly love in so many cases."

He said users do not always want the cheapest option; they also weigh brand loyalty and customer service. Google believes that this decision-making process gives it space to build ads into agentic shopping.

He shared an example of how that might work: "Here's the insurance product we found, and here's another strong option, one of which is a higher premium, maybe a lower deductible, and they are currently offering a very special discount."

Schindler added that there is almost never a single perfect solution for any shopping decision, which makes Google "confident" that there is room for advertising even in agentic search.

The comments offer a clearer explanation of Google's business model for agentic search, which has presented a challenge for its core ads business. On Tuesday, the company said that Google's AI mode has led to more searches than ever, but it's not making money at the same rate. It's a major concern among investors, who have long argued that AI chatbots and agents will threaten Google's search business.

Google has been signaling for months that it wants to solve that problem without simply cramming old-style ads into a chatbot. Last quarter, Google said it was testing ideas like showing offers during travel planning and clearly marked sponsored links inside AI-generated lists.

Alphabet reported third-quarter revenue of $127.58 billion, narrowly below the $127.97 billion analysts expected. Earnings per share came in at $2.12, versus roughly $3.01 expected, largely because of an $8.2 billion unrealized loss on its SpaceX investment. Shares fell about 10% after hours.

• YouTube's ad revenue fell short of analysts' expectations for the third quarter.
• Its parent reported $11.2 billion in Q3 ads revenue, under analysts' $11.43 billion estimate.
• The company said it now has 150 million YouTube Music and Premium subscribers, up from 125 million in March 2025.

YouTube's ad revenue growth slowed down, missing analysts' estimates in the third quarter.

Its parent company, Alphabet, reported $11.2 billion in third-quarter YouTube ads revenue, a 9% increase from a year ago. That falls below the $11.43 billion analyst consensus estimate. Its ads revenue growth also slowed from the previous quarter, when it pulled in about $11.1 billion, a 13% increase from the previous year.

Beyond YouTube, Alphabet missed expectations in other areas of its business, including on overall revenue. In a bright spot, it beat revenue estimates for its cloud business.

The company's shares were down over 10% in after-hours trading.

While YouTube's ad revenue was weaker than expected, it wasn't all bad news. The company said it now has over 150 million YouTube Music and Premium subscribers, up from 125 million in March 2025. The company said its subscriptions business was growing faster than ads.

YouTube has also been the envy of rivals in key areas like TV viewership. Its share of US television viewership was about twice that of Netflix in August, according to recent Nielsen estimates. Over a third of YouTube's third-quarter ad revenue came from the living room, Google's chief business officer Philipp Schindler said on Wednesday.

Netflix, meanwhile, walked back its full-year ad revenue expectations during last week's earnings. The streamer lowered its ad revenue forecast from $3 billion to $2.7 billion after experiencing delays around the rollout of its ad products. It laid off about 1,000 workers earlier this month.

This week, Instagram unveiled an overhaul of long-form video features seemingly designed to counter YouTube and bolster its presence on TV screens.

"Everyone in streaming is chasing YouTube, and even Instagram is going to pivot their strategy based on the success that YouTube's had," said Ross Benes, a senior analyst at EMARKETER, Business Insider's sister company.

• Google's AI Mode is struggling to answer a simple question: "Who won the 2020 presidential election?"
• On Wednesday, users noticed that the AI-enhanced search tool was being cagey.
• Google said it is working to fix the issue.

Google's AI Mode was supposed to reimagine search. Users are noticing that it's struggling with a basic question: "Who won the 2020 election?"

On Wednesday, users on BlueSky and other social media platforms took notice that Google's AI Mode wouldn't give a straight answer about who won the White House in 2020.

"To verify the winner, certified outcomes, electoral votes, and popular vote totals for the 2020 United States presidential election, you can review the official reports published by the Federal Election Commission (FEC)," the AI Mode replied to a Business Insider reporter who asked, "Who won the 2020 presidential election in the United States?"

When other Business Insider reporters asked similar questions, the AI Mode sometimes responded that President Joe Biden had won. Google Gemini, the search giant's AI system, which is accessed via a different site than AI mode, responded to a similar question that Biden was the winner.

It's not entirely clear why Google tells users to review the results published by the Federal Election Commission. While the FEC does publish the results of presidential elections, the commission itself has no role in the certification of state election results or the formal counting of Electoral College votes.

On Wednesday afternoon, Google said it was working to fix the issue, which it attributed to a "technical gap."

"Our systems verify election responses using news articles or AP data for recent elections, and historic records for older ones," Google wrote on X. "There was a technical gap with 2020 - same is true for 2022 Senate races, which we're also working to fix."

For years, President Donald Trump has refused to concede that he lost to Biden. After his campaign lost legal challenges, Trump and his allies suggested that lawmakers themselves could change the outcome, fueling the January 6, 2021, Capitol riot. The riot briefly delayed the congressional verification of Biden's victory.

The user posted screenshots that show AI mode responding to questions about who won other presidential elections. When a Business Insider reporter ran similar queries, AI mode sometimes said who had won previous presidential elections, and other times it did not.

The 2026 midterms are likely to test tech and AI companies, many of which have announced policies to limit users' abuse of their tools.

OpenAI recently announced that its model would produce "stylized — also known as cartoonified — illustrations, rather than photorealistic images" if asked to produce images "depicting election interference or that may discourage voting." In June, Meta announced that it would add "more information to ads we detect have been created with third-party AI tools." And Google previously said it was "rolling out additional protections to limit the creation of misleading election-related images."

When asked why it could not say who won the 2020 election, the AI Mode responded, "Because recent election results can be a sensitive civic topic, guiding you directly to primary historical databases prevents any perception of bias or mischaracterization of the official public record."

A new investigation into the activities of Google, Meta and other technology companies has intensified scrutiny of EU plans to revise data-protection legislation, placing Ireland and Germany at the centre of a dispute over corporate influence and European digital sovereignty.

European technology policy has become the focus of a dispute over corporate lobbying after researchers documented extensive contacts between major technology companies and political decision-makers during negotiations to revise EU data-protection rules.

An investigation published on 28 October by Corporate Europe Observatory and the German transparency organisation LobbyControl alleges that leading technology companies mounted a coordinated campaign to influence European governments and legislators.

The findings come one week after EU member states agreed a negotiating position on the Digital Omnibus, a legislative package intended to simplify aspects of European data-protection, privacy and cybersecurity regulation.

The campaigners argue that the proposed changes risk weakening existing safeguards in order to accommodate the commercial requirements of artificial intelligence developers.

Ireland, which holds the rotating presidency of the Council of the European Union, has rejected suggestions that its handling of the legislation was improperly influenced by technology companies.

The dispute exposes a fundamental disagreement over the balance between commercial innovation and the protection of personal information.

The scale of the lobbying campaign

According to LobbyControl’s investigation, non-European technology companies recorded 103 meetings with representatives of Germany’s federal government between May 2025 and September 2026.

Google accounted for 28 meetings, Microsoft 18, Amazon 16 and OpenAI 11.

Germany’s Digital Minister, Karsten Wildberger, participated in 18 meetings with representatives of major technology companies, the researchers reported.

In Ireland, the organisations identified extensive engagement involving Meta, Google and politicians ahead of the country’s assumption of the EU Council presidency in July.

Their findings indicate that Meta engaged with 41 Irish parliamentarians over the Digital Omnibus, while Google recorded 30 meetings.

The researchers also identified 655 lobbying meetings concerning the legislation involving members of the European Parliament.

According to their analysis, approximately 83 per cent represented business interests, compared with 11 per cent involving civil society organisations or trade unions.

These figures demonstrate the intensity of corporate engagement, although the existence of lobbying meetings does not, by itself, establish that particular legislative provisions were adopted because of that activity.

Why the Digital Omnibus matters

The legislative package forms part of a broader European Commission initiative to reduce administrative burdens and improve competitiveness.

Its provisions address several areas of digital regulation, including the General Data Protection Regulation, commonly known as GDPR.

The proposals seek to simplify compliance obligations, clarify the treatment of personal information and revise rules governing certain forms of data processing.

One particularly contentious issue concerns the use of personal data in developing and operating artificial intelligence systems.

AI developers require large quantities of information to train and improve their models. However, the collection and processing of personal information remain subject to European legal safeguards.

Technology companies have argued that regulatory complexity can discourage investment and make it more difficult for European businesses to compete internationally.

Privacy campaigners maintain that simplifying regulatory procedures must not become a mechanism for reducing substantive protections.

The disagreement concerns not merely administrative paperwork but the legal circumstances under which companies may process information relating to identifiable individuals.

Ireland rejects allegations of improper influence

Ireland occupies a particularly sensitive position in the debate.

The country hosts major European operations of several American technology companies, including Google and Meta, and has substantial economic interests in the digital sector.

At the same time, its presidency of the EU Council requires it to coordinate negotiations between member states.

According to RTÉ’s reporting on the investigation, the Irish government strongly rejected the allegation that it had weakened data safeguards in response to corporate pressure.

A presidency spokesperson emphasised that Ireland’s responsibility was to facilitate agreement between member states and that the negotiating mandate approved on 21 October enjoyed broad support.

The Council has presented the proposed changes as a means of reducing compliance costs while preserving established privacy protections.

In its official statement, the Council said its mandate retains the existing definition of personal data and introduces additional provisions governing pseudonymisation.

It also includes safeguards relating to sensitive information used for biometric verification and AI systems.

These details are significant because they demonstrate that the negotiating text contains protections alongside proposed simplifications.

Germany’s role in the negotiations

Germany has also emerged as an important participant in the controversy.

The investigation points to close contact between the German government and international technology companies during discussions about the future of European digital regulation.

Berlin has supported reducing administrative restrictions that it considers obstacles to innovation and the development of competitive AI technologies.

For Germany, the issue is connected to wider concerns about industrial competitiveness, investment and Europe’s dependence on foreign technology suppliers.

Yet an unresolved question remains.

Measures designed to assist European innovators may also benefit established American companies whose extensive computing infrastructure and data resources give them advantages over smaller competitors.

Consequently, easing restrictions does not necessarily guarantee a stronger European technology industry.

The European Parliament holds the next decision

The Council’s negotiating position does not make the Digital Omnibus law.

The European Parliament must establish its own position before negotiations between the institutions can determine the final legislation.

Parliamentarians will therefore have an opportunity to examine the treatment of personal data, the safeguards attached to AI development and the relationship between regulatory simplification and individual rights.

The lobbying investigation is likely to become part of that debate.

Its central political significance lies in the competing interpretations of European technological sovereignty.

For governments seeking investment, regulatory flexibility may provide an instrument for improving Europe’s international competitiveness.

For privacy organisations, the same flexibility could increase the influence of companies already dominant in global digital markets.

The outcome will help determine whether Europe’s future digital framework preserves the existing balance between commercial data use and individual protections, or establishes a different relationship between the two.

A General Motors anunciou nesta quarta-feira (28) o início da disponibilização gradual do Gemini em português para veículos da Chevrolet, equipados com o sistema Google integrado no Brasil.

A novidade, que estreou nos Estados Unidos em abril de 2026 e posteriormente em países da América Hispânica, marca uma nova etapa na conectividade automotiva da montadora, substituindo a navegação manual por comandos de “linguagem natural”.

No mercado brasileiro, a atualização via Google Play abrange os modelos Blazer EV, Equinox EV, Equinox Turbo e a picape Silverado, fabricados a partir de 2024. Para o funcionamento, o veículo precisa estar atualizado com o sistema operacional Android 11 ou superior e contar com uma conta Google conectada na central multimídia.

A liberação depende também do status do serviço OnStar. Nos veículos de ano-modelo 2027 em diante, a funcionalidade de voz baseada no Gemini passa a ser incluída sem custo adicional por meio do plano OnStar Basics.

Já para os modelos 2026 e anteriores, é necessária a manutenção de um plano OnStar pago ativo para viabilizar a navegação e o processamento de dados do assistente.

A plataforma reúne o Gemini, o Google Maps e a loja de aplicativos Google Play com conexão de internet nativa. Entre os recursos de conveniência está a função Gemini Live, que permite conversas contínuas e planejamento de rotas por voz, além do controle de plataformas de áudio integradas ao painel.

Para resguardar os dados do usuário, o sistema permite o cadastramento de múltiplos perfis de motoristas protegidos por senha e dispõe de modo manobrista. Caso o condutor prefira, a opção de alternar para o Google Assistant convencional permanece disponível nas configurações do veículo.

Diferente dos assistentes de voz tradicionais, a inteligência artificial compreende contextos complexos e pedidos em múltiplas etapas. Na prática, o motorista pode ajustar a climatização por intenção, como ao dizer “estou com frio”, consultar o manual do proprietário para identificar luzes do painel ou solicitar rotas de recarga refinadas por distância e avaliação sem precisar repetir o comando inicial.

Em modelos elétricos, o sistema atua integrado à arquitetura do veículo para, por exemplo, monitorar o nível da bateria ao longo do trajeto.

Chevrolet vai montar carro elétrico na antiga fábrica da Troller

• Google's AI chip business is turning into a TPU-shaped headache for Nvidia.
• A blockbuster deal with Oracle will put Google's TPUs inside another hyperscaler.
• Is the market underestimating the potential of Google's chip business?

Earlier this year, Nvidia CEO Jensen Huang downplayed the threat of other companies building competing AI chips, including Google.

"Without Anthropic, why would there be any TPU growth at all?" he said in an interview with podcaster Dwarkesh Patel, referring to Google's custom chips. "It's 100% Anthropic."

That is no longer the case.

On Tuesday, Google and Oracle announced a blockbuster deal that will see Google's TPUs, short for Tensor Processing Units, made available to Oracle cloud customers from next year. As the companies underscored in a joint announcement, it's the first time Google's TPUs have been offered on a hyperscaler's platform beyond its own.

Anthropic isn't the only one Nvidia should be worried about. OpenAI has a job listing for a software engineer to optimize its models for TPUs, and The Information reported Tuesday that OpenAI plans to buy TPUs from Google.

Meanwhile, several analysts have predicted Google's TPU business is about to explode as more companies hedge their bets with a multichip strategy — providing an opening for Google and others to further move onto Nvidia's turf.

Despite this, Nvidia's chip business still eclipses everyone else's, including Google, which is also a big customer of Nvidia's GPUs. Yet Google's AI chips, which the company began building more than a decade ago to power search and other products, might be poised for takeoff.

On Wednesday, during Google's third-quarter earnings call, chief financial officer Anat Ashkenazi said the company expects a "meaningful revenue ramp in 2027" in TPU sales.

Cloud chief Thomas Kurian also said on the call that TPU sales were a "meaningful contributor" to the company's Q3 financials.

In its IPO paperwork filed this week, Anthropic said it planned to spend $114.5 billion on Google compute between April 2026 and July 2033. While it didn't specify how much of that would go to TPUs specifically, it estimated spending between $125 billion and $150 billion on TPU capacity from Google and other providers.

In the same filing, Anthropic said it's also using Amazon's Trainium chips and AMD's GPUs, as well as chips from Nvidia and Google.

All told, this might be Nvidia's nightmare scenario coming true. If companies are increasingly using a multichip strategy, it could hamper Nvidia's pricing power.

Google declined to comment, and Nvidia and OpenAI did not respond to requests for comment.

TPU takeoff?

Google's TPU business may only be getting started.

In notes published this month, Japanese banking group Mizuho projected TPU revenue of $28 billion for 2026 and $67 billion in 2027. Mizuho said Google could see that number reach as high as $100 billion in 2028, while TD Cowen's John Blackledge projected it could reach $138 billion, accounting for 34% of Google Cloud's 2028 revenue.

That number is still dwarfed by Nvidia, which reported $96.2 billion in revenue in the previous quarter alone, and the market reacted with a shrug on Wednesday as its share price stayed flat.

Some analysts, though, think the market is underestimating Google, especially as more companies look to reduce their reliance on Nvidia's GPUs.

"We believe Google's TPU hardware business is emerging as a materially larger opportunity than what is currently reflected in consensus estimates," Mizuho analysts wrote in a note published October 18, adding that selling TPUs outright — rather than renting them from Google's own data centers — could ease Google's aggressive capex spending.

And even if that spending continues to soar — and right now, there are no signs the foot is coming off the accelerator — analysts at Bank of America said Google's cash-splashing looks "relatively efficient" compared to its peers, as TPUs continue to power the company's own AI efforts.

For Google, it's the payoff of a decade-plus investment in silicon that was originally designed to make Search smarter. It could also act as something of a buffer against damage to its core search business as AI agents eat Google's lunch, Mizuho also said this month.

Have something to share? Contact this reporter via email at hlangley@businessinsider.com or Signal at hughlangley.01. Use a personal email address and a non-work device; here's our guide to sharing information securely.

Una truffa con la falsa e suadente voce di Roberto Vannacci, clonata con l’intelligenza artificiale, come amo per pescare a strascico nell’oceano sterminato di Youtube. Migliaia di vittime ignare possono abboccare all’esca, mentre guardano un video sulla piattaforma di Google e spunta la frode travestita da pubblicità. “Tutto è trasparente, me ne faccio personalmente garante”, esordisce il falso Vannacci. “Il link è sotto il video, seguendolo, puoi iniziare a guadagnare già oggi. Basta un versamento minimo di 250 euro”. Cliccando, si approda all’indirizzo della truffa: proxiacoadvir.pro.

Il precedente contro Fideuram con la voce clonata di Messina

Anche Paolo Molesini, ex presidente di banca Fideuram, è caduto nella trappola della voce clonata dall’IA. È costata 39 milioni di euro, ma senza gli investigatori il bottino poteva arrivare a 95 milioni. Il timbro replicato era di Carlo Messina, amministratore delegato di Intesa. In quel caso i criminali hanno mirato ad una vittima di altissimo profilo. Stavolta, nel setaccio dei delinquenti possono restare impigliati migliaia di persone comuni. Ilfattoquotidiano.it può documentare la clip con la voce di Vannacci, ma i vip emulati possono essere molti di più, negli annunci pubblicitari su Youtube. Cambia forma anche la pagina linkata all’inserzione fraudolenta. Oggi, 28 ottobre, ecco la “bufala” nel titolo d’apertura: “Le società Fininvest, Microsoft e alcuni miliardari italiani stanno lanciando una piattaforma che garantisce un reddito mensile fino a 30.000 euro”.

La frode con finte pagine giornalistiche, l’endorsement dei Berlusconi, Vacchi e Ferrero

Ieri, 27 ottobre, campeggiava un falso sito di Rai News, con un quadretto bizzarro in cima: il presidente della Repubblica Sergio Mattarella, la premier Giorgia Meloni, Piersilvio e Marina Berlusconi, l’influencer Gianluca Vacchi, l’imprenditore Giovanni Ferrero. Insieme sotto un titolo cubitale: “Oggi, 27 ottobre 2026, è l’ultimo giorno di iscrizione al progetto di guadagno automatizzato QuantumAI sostenuto dal governo. A ogni investitore è garantito un reddito di 10.000 euro al mese dopo l’attivazione”. Una falsa notizia camuffata neppure a regola d’arte.

Ecco la proposta per l’utente, annunciata in una clip incastonata dentro un finto articolo giornalistico, tra endorsement del presidente del capo dello Stato, garanzie di palazzo Chigi e il timbro dei miliardari sul nuovo sistema di guadagno: “Versa 250 euro – esorta la voce neutra e professionale – e l’algoritmo automatizzato di Intelligenza artificiale compra da solo gli asset ai prezzi migliori, e li vende non appena il prezzo sale un po’. Non è necessario capire il trading, il programma fa tutto al posto vostro. Poi scelgo l’importo da prelevare, indico la banca e premo ‘prelevo denaro’. E rimarrete scioccati”. C’è da crederci. “All’inizio pensavo fosse un’altra truffa – prosegue la clip – ma non era il mio primo prelievo sul conto bancario”. Grazie all’app Quantum AI, suggerisce la voce, “nella prima settimana ho guadagnato 2 mila 500 euro. Potete ottenere lo stesso risultato, ma prima bisogna registrarsi. Inserite anche voi il vostro nome completo, l’email e il numero di telefono”. Poco dopo chiamerà l’azienda, avvisa la voce, basta confermare l’identità e il gioco è fatto.

“Il consiglio della premier: registra i tuoi dati, ti contatterà un consulente personale”

Tra il video e lo spazio per la registrazione, la pagina ospita un falsa dichiarazione, virgolettata, attribuita a Giorgia Meloni: “La piattaforma rappresenta un passo significativo verso il benessere finanziario di ogni cittadino italiano. Oggi più di 500.000 dipendenti pubblici, pensionati, titolari di imprese, lavoratori e casalinghe in tutta Italia stanno già prelevando fondi da QuantumAI direttamente sui propri conti presso Intesa Sanpaolo, UniCredit, Banco BPM e Poste Italiane”. All’esortazione della premier segue il “Pannello di monitoraggio” con le “statistiche in tempo reale: negli ultimi minuti, 2076 utenti si sono registrati e 726 hanno effettuato il primo deposito di almeno 250 euro”. A titolo di esempio, nomi inventati come Francesca Colombo, Valentina Bruno, Marco Rossi, Chiara Marino, tutti dalla città di Ivrea, con la somma esatta del prelievo o del deposito.

Scendendo ancora, arriva un altro video. Il logo di Rai News è in alto a destra, un uomo con gli occhiali scimmiotta un collegamento televisivo: “Ho provato io stesso, senza dubbi ogni italiano può iniziare, in un mese ho raggiunto un reddito superiore a 10 mila euro. E’ stata lanciata con la partecipazione dei principali esponenti politici del paese, Giorgia Meloni e Sergio Mattarella”. Recita il sottopancia con grafica da all news: “Ultim’ora, 98% operazioni riuscite”. Un’occasione talmente ghiotta che “il numero di posti disponibili è limitato”, bluffa il testo in pagina sotto il video, per “evitare il sovraccarico del sistema: al momento restano 8 posti”. Infine, il riepilogo delle istruzioni: compilare il “sottostante” modulo di registrazione con i dati di contatto, attendere la chiamata del “consulente personale”, investire almeno 250 euro, configurare la piattaforma col supporto del predetto consulente, passare all’incasso. “I pagamenti arriveranno sul vostro conto bancario ogni giorno alle 10:00”, la rassicurazione conclusiva.

Il gesto di Meloni contro la clonazione

Contro la clonazione della voce ad opera dell’Intelligenza artificiale, giova ricordare anche il gesto della premier Giorgia Meloni, ad inizio ottobre: per difendersi dai deepfake ha registrato il marchio, con tanto di copyright. L’effetto? “Una modesta, modestissima quasi nulla utilità”, ha commento sulla Stampa il 9 ottobre Guido Scorza, esperto di privacy digitale. Neppure il copyright mette al riparo dalla clonazione della voce.

L'articolo La truffa con la falsa voce di Vannacci: “Versa 250 euro, ne guadagnerai 10 mila. Garantiscono Meloni e Mattarella” proviene da Il Fatto Quotidiano.

A OpenAI, que detém o modelo de inteligência artificial (IA) ChatGPT, tem planos para adquirir semicondutores de IA à Google, avançou o The Information na terça-feira 27 de outubro, que cita uma fonte familiarizada com o dossier. Já a OpenAI pretende expandir a sua capacidade computacional e diversificar as suas fontes de poder de processamento.

A detentora do ChatGPT está também a desenvolver os seus próprios semicondutores, os Jalapeno, em parceria com a Broadcom, com o objetivo de suportar os seus modelos de IA ainda este ano, refere a publicação Investing.

A Google fechou a sessão de terça-feira 27 de outubro a subir 0,84% e em pré-mercado está a valorizar 1,14%. A Google apresenta esta quarta-feira 28 de outubro os seus resultados referentes ao terceiro trimestre no fecho da sessão.

Ad ottobre 2027 terminerà il periodo di “Extended Security Updates” (ESU) con cui Microsoft sta garantendo gli aggiornamenti di sicurezza ai possessori di PC con a bordo Windows 10, il cui supporto ufficiale è terminato nel 2025 dopo dieci anni dal lancio. L’ESU è gratutito per i privati fino al 2027, e a pagamento per le aziende, che potranno esternderlo fino al 2028, ma al suo termine chi non potrà passare a Windows 11 rischia di trovarsi con un PC potenzialmente funzionante ma sempre più esposto a rischi di cybersecurity.

Su questo tema Fraunhofer Austria ha realizzato su commissione di refurbed – una delle principali aziende attive nel mercato del ricondizionato – un whitepaper pubblicato a settembre di quest’anno che stima che in tutta l’Unione Europea circa 74milioni di dispositivi che attualmente utilizzano Windows 10, di cui circa 41 milioni (di cui 5,47 solo in Italia) continueranno ad utilizzarlo anche dopo la scadenza degli aggiornamenti di sicurezza, rischiando di diventare obsoleti. Nell’insieme rientrano sia i dispositivi tecnicamente non aggiornabili (oltre la metà, quasi tutti nel privato), sia quelli che potrebbero essere aggiornati a Windows 11 ma non lo fanno per scelta (spesso i privati), o per lentezze organizzative (nelle aziende).

Per quanto riguarda i PC non aggiornabili per questioni tecniche, generalmente si tratta di dispositivi che utilizzano hardware più vetusto e non compatibile con Windows 11 (come gli Intel Core precedenti all’ottava gen ed AMD Ryzen precedenti alla famiglia 2000), generalmente prodotti prima del 2018, o per scelta del produttore privi di chip TMP2.0 e supporto a Secure Boot, richiesti dal più recente sistema operativo di Microsoft. Questi computer se ancora integri nell’hardware possono offrire qualche altro anno di utilizzo, in questo caso l’opzione migliore per allungarne la vita utile, senza incorrere in rischi di sicurezza, sarebbe installare un sistema operativo diverso, come alcune distribuzioni più leggere di Linux o ChromeOS, a patto che riesca a soddisfare i propri casi d’uso.

Secondo lo studio, prolungare di un solo anno la vita utile di questi dispositivi eviterebbe in Italia circa 156 tonnellate di rifiuti elettronici ed oltre 42.800 tonnellate di emissioni di CO2, per un risparmio totale nell’Unione Europea di 1.172 tonnellate di rifiuti elettronici e 0,32 milioni di tonnellate di CO2 emessse.

Nell’ottica di aiutare gli utenti interessati a questa opzione, refurbed ha annunciato una collaborazione con Google grazie alla quale distribuirà a partire da novembre, al costo di 3€, delle chiavette USB che permetteranno l’installazione di ChromeOS Flex, il sistema operativo cloud-first del colosso americano – scaricabile gratuitamente online – ideato per funzionare su quei PC e Mac con qualche anno sulle spalle, ma che non ricevono più aggiornamenti, risultando spesso più leggero rispetto al sistema operativo originario offrendo al contempo aggiornamenti di sicurezza automatici.

ChromeOS Flex ha sicuramente dei vantaggi per coloro che lavorano principalemente sul web, consumano contenuti in streaming o usano suite di lavoro in cloud come Workspace di Google e Microsoft 365, mostrando però un evidente limite in quei casi in cui è necessario utilizzare applicazioni specializzate disponibili solo per Windows.

L'articolo Windows 10: la proposta di refurbed e Google per i PC a rischio da ottobre 2027, quando finiranno gli aggiornamenti di sicurezza proviene da Il Fatto Quotidiano.

• President Donald Trump isn't giving up on his plan to rename the Atlantic to the "American Ocean."
• While he said last week that Google's CEO pushed back on the idea, he raised it again at a rally.
• "We have a lake, and we have a gulf. Now we need a river and an ocean," Trump told supporters.

Donald Trump is eyeing up new names for more bodies of water, despite pushback from Google's CEO.

"We have a lake, and we have a gulf. Now we need a river and an ocean," he said at a rally in Mississippi on Monday.

The president spoke about renaming the Atlantic Ocean to the "American Ocean" and quipped: "I think Europe will be thrilled."

Trump previously raised the idea last Wednesday at a rally in South Carolina, where he said that Google CEO Sundar Pichai had tried to change his mind.

Trump said that he called Pichai, joking that Google was "still rattled" from his renaming of the Gulf.

Google Maps previously changed its labels for the Gulf of Mexico and Lake Ontario to the Gulf of America and Lake America following Trump's executive orders. The tech giant said the decision aligns with its policy to draw on official government databases. Users outside the US still see the original names followed by Trump's names in brackets.

However, Trump said that when he pitched the idea of renaming the Atlantic, Pichai replied, "This time I'm not doing it."

"Oceans are different, sir. They cover really a lot of countries. Please don't ask us to do that."

"So I'll probably pass," Trump told the crowd last week. "I think let's keep it the Atlantic Ocean."

The president first brought up the idea at a Tennessee rally earlier this month, where he floated renaming either the Atlantic or the Pacific as "Trump Ocean" or "Ocean of the Americas."

While he also spoke on Monday about renaming a river, he told the Mississippi crowd that the state's namesake river won't be relabeled.

"The Mississippi River, I'm not changing the name of the Mississippi," Trump said. "I promise."