Brazil Splitting AI Infrastructure Between Superpowers

The Brazilian government has announced a strategic move to bolster its domestic artificial intelligence capabilities by launching a significant supercomputer infrastructure project. Rather than aligning exclusively with a single geopolitical bloc, Brazil has opted for a hybrid procurement strategy that distributes key technological responsibilities between Chinese and American corporations. This decision reflects a careful balancing act, positioning the nation to leverage diverse hardware architectures and software ecosystems while navigating the intensifying technological rivalry between Beijing and Washington.

Imagen generada con IA

The project, overseen by Brazil’s Ministry of Science, Technology and Innovation, aims to provide high-performance computing resources necessary for training large-scale machine learning models and advancing scientific research. By diversifying its partnerships, Brazil intends to avoid vendor lock-in and ensure that its digital infrastructure remains resilient to potential supply chain disruptions or export control fluctuations. The initiative is particularly focused on enabling local researchers and private firms to develop AI models tailored to regional datasets, such as those related to the Amazon rainforest, agriculture, and public health.

For the technical community, this development is significant because it highlights the growing necessity of building compute-heavy infrastructure that is not beholden to a single nation's geopolitical influence. By integrating hardware from multiple sources, Brazil is essentially creating a heterogeneous computing environment that may require sophisticated orchestration layers to manage workloads effectively. This setup poses unique challenges in software compatibility and kernel-level optimizations, as engineers will need to ensure that AI training pipelines can move seamlessly between different hardware backends.

Ultimately, this initiative underscores the reality that nations seeking AI sovereignty must often navigate a fragmented global tech landscape. By splitting these contracts, Brazil is signaling that it prioritizes technical functionality and infrastructure accessibility over exclusive strategic alignment. Whether this dual-path strategy will result in a cohesive national compute strategy or lead to complex integration bottlenecks remains a pivotal question. However, it represents a notable blueprint for emerging economies attempting to participate in the global AI race without sacrificing diplomatic neutrality or compromising their long-term digital independence.

Artículos relacionados de LaRebelión:

• US Lawmakers Investigate Chinese AI in American Firms
• Chinese AI Firms Copying Claude Millions of Queries Exposed
• Chinese AI Models Spur Corporate Safety Dilemma
• Nvidia Launches Its First Open-Source AI Model
• Anthropics AI Launches Rogue Malware Attack on GitHub

---

Fuente Original: Reuters

Artículo generado mediante AI.larebelion.

AI Exploits Threaten US Critical Infrastructure PLCs

The cybersecurity landscape is facing a new and worrying threat: AI-generated exploit scripts specifically designed to target Siemens S7 PLCs. These programmable logic controllers are crucial components in the operational technology (OT) systems that underpin much of the United States' critical infrastructure, including energy grids, water treatment facilities, and manufacturing plants. The development of such sophisticated and tailored attack tools by artificial intelligence raises significant concerns for national security and public safety.

The danger lies in the ability of AI to automat e and accelerate the creation of malware. Previously, developing exploits required significant human expertise and time. Now, AI can rapidly identify vulnerabilities in complex industrial control systems and then generate the precise code needed to take advantage of them. This democratisation of advanced hacking tools means that even less sophisticated malicious actors could potentially launch devastating attacks against vital infrastructure. The report highlights that these AI-driven exploits could bypass existing security measures, leading to potential disruptions in essential services. The speed at which these exploits can be developed and deployed poses a formidable challenge to cybersecurity professionals tasked with defending these systems.

This development underscores the urgent need for enhanced security measures specifically for OT environments. Traditional IT security approaches may not be sufficient to counter these AI-powered threats. Companies and government agenc ies responsible for critical infrastructure must prioritise robust cybersecurity strategies, including continuous monitoring, rapid patching, and the implementation of advanced threat detection systems that can identify and neutralise AI-generated exploits before they can cause harm. The proactive defence against these evolving threats is paramount to safeguarding the nation's essential services from disruption.

Fuente Original: https://thehackernews.com/2026/08/ai-generated-exploit-scripts-target.html

Artículos relacionados de LaRebelión:

• OpenAI Halts Astra AI Over Critical Cybersecurity Risks
• Russia Exploits Doorbell Cameras Targeting NATO Infrastructure
• Google Awards 250K for Critical Linux VM Escape
• AI Coding Agents Vulnerable to Bash Exploits
• Enterprise AI Faces Critical Prompt Injection Threats

Artículo generado mediante LaRebelionBOT

Grok CLI Archivos Subidos a la Nube Sin Permiso

La reciente introducción del modelo Grok 4.5 de xAI, accesible a través de su interfaz de línea de comandos (CLI) Grok Build, ha generado una gran controversia. Varios desarrolladores notaron un comportamiento alarmante: la CLI estaba subiendo de forma predeterminada todos los archivos locales de sus directorios de trabajo al servicio de almacenamiento en la nube de xAI, sin previo aviso ni consentimiento explícito.

Investigaciones independientes revelaron que Grok CLI transmitía el contenido de archivos, incluyendo secretos sensibles como archivos `.env`, directamente a xAI. No solo esto, sino que subía el repositorio completo, incluyendo el historial de Git, independientemente de los archivos que el a gente realmente necesitara leer. Estos datos se almacenaban en un bucket de Google Cloud Storage llamado `grok-code-session-traces`, y esta funcionalidad estaba activa por defecto, sin poder desactivarla fácilmente.

La situación fue calificada como grave por varias razones. Enviar el código fuente completo a un servidor de LLM no es una práctica estándar; las alternativas más seguras implican indexar el código localmente y enviar solo metadatos o embeddings. Además, la transmisión sin cifrar de archivos `.env` es una brecha de seguridad inaceptable, dado que contienen credenciales y claves de acceso críticas. La falta de notificación a los desarrolladores sobre esta recopilación masiva de datos provocó indignación generalizada.

Ante la reacción pública, xAI deshabilitó la función de carga de archivos mediante una bandera remota. La respuesta oficial de SpaceX fue criticada por ser condescendiente y culpar implícitamente a los usuarios. Sin embargo, baj o presión, incluyendo comentarios de figuras como Sam Altman, SpaceX se comprometió a hacer de código abierto la CLI de Grok y a eliminar los datos recopilados. Si bien el código abierto se lanzó apresuradamente y con errores, representa un paso hacia una mayor transparencia.

Este incidente pone de manifiesto la importancia de la confianza en los proveedores de herramientas de desarrollo. La secreta y masiva carga de datos de Grok CLI ha erosionado la confianza de la comunidad. Para reconstruirla, SpaceX/Grok deberá demostrar un compromiso a largo plazo con la privacidad y la seguridad, y probablemente ofrecer incentivos significativos para que los desarrolladores vuelvan a confiar en su producto.

Fuente Original: https://blog.pragmaticengineer.com/grolk-cli-uploaded-all-your-files-to-the-cloud/

Artículos relacionados de LaRebelión:

• IA Identidad Segura Pero sin Control
• Grok Bot: Tus Nuevos Compañeros de Trabajo Digitales
• Memoria IA Tencent Comparte Contexto de Equipo Sin Gobernanza
• IA Potente en Raspberry Pi sin GPUs
• IA Nuevos Riesgos de Privacidad Sin Fugas de Datos

Artículo generado mediante LaRebelionBOT

AWS Pledges $500M to Cloud and AI Education

Amazon Web Services has announced a commitment of more than $500 million toward training students in cloud computing and artificial intelligence skills. The initiative represents one of the larger corporate investments in technology education seen in recent years, and it signals how seriously major cloud providers are treating the growing gap between AI talent demand and supply.

Imagen generada con IA

The program is designed to reach students across a broad range of educational levels and backgrounds, with the goal of equipping a new generation of workers with practical, job-ready skills in AWS cloud services and AI tooling. This kind of initiative matters because the shortage of qualified cloud and AI practitioners is a genuine bottleneck for enterprises trying to accelerate their digital transformation efforts. By funding education at scale, AWS is effectively investing in its own future customer and developer base while simultaneously addressing a real workforce problem.

AWS has an existing infrastructure for this type of outreach through its AWS Educate and AWS Academy programs, which already partner with educational institutions to embed cloud training into curricula. The new funding is expected to significantly expand the reach and depth of these offerings, potentially adding new certifications, hands-on lab environments, and scholarship opportunities for learners who might not otherwise have access to this kind of training.

From a competitive standpoint, this move also makes strategic sense. Microsoft and Google have both made significant pushes into AI education through their own cloud platforms, and AWS cannot afford to cede that pipeline of future developers and architects. Developers trained on a particular cloud ecosystem tend to advocate for and build on that platform professionally, making early education a long-term retention and adoption play.

The broader context here is the rapid acceleration of generative AI adoption across industries. Businesses are eager to deploy AI-powered applications but frequently lack the internal expertise to do so effectively. A larger pool of AWS-trained developers and data practitioners directly benefits the companies that rely on those workers, and by extension strengthens the AWS ecosystem as a whole.

For technically literate readers watching the cloud market, this announcement is less of a philanthropic gesture and more of a calculated infrastructure investment. Building the human capital layer around a cloud platform is just as important as expanding data center capacity or launching new services. AWS appears to understand that the next phase of cloud growth depends as much on skilled people as it does on hardware and software.

Artículos relacionados de LaRebelión:

• Gigantes Tecnologicos IA Impulsa Cloud Google y Amazon Lideran
• OpenAI Halts AI Safety Training A Bold Step
• Developer Bridges macOS and Obscure Printer Gap
• Amazon Escanea Libros Raros para IA Destruccion de Cultura
• AIs Accuracy Trick Modules Cheat Training Role Anchor Fixes It

---

Fuente Original: About Amazon

Artículo generado mediante AI.larebelion.