📁 Altro

As robots and autonomous vehicles speed toward market, safety frameworks struggle to keep pace. Integrating Large Language Models into physical systems introduces novel risks. For those managing on-premise and edge deployments, the challenge is twofold: ensuring low latency and data protection without compromising reliability.

LG's R&D division at Sciencepark starts talks with SpaceX while targeting tangible results by 2030. The move signals a convergence between space infrastructure and artificial intelligence, where autonomous and low-latency computing demands drive local LLM deployment in extreme on-premise settings, with direct impact on edge hardware and data sovereignty.

The ongoing climate event threatens grid stability, while AI model inference and training push demand to record levels. In this scenario, fuel cells emerge as an option for those managing on-premise infrastructure. AI-RADAR examines the convergence.

Kiwibit’s smart feeder uses on-device AI to identify birds, turning backyard birdwatching into a game. Beneath the playful concept lies a real exercise in edge inference on constrained hardware, highlighting the optimization challenges familiar to on-premise system designers.

A security experiment shows how brittle the AI agent ecosystem can be: a fake skill, pushed via an Instagram ad, bypassed every scanner and reached over 26,000 agents, including corporate accounts. The incident raises hard questions about AI software supply chains and the risks for organizations that embrace agents without direct control over the infrastructure.

Ten years after the vote, a landmark study puts the cost of leaving the EU at 6–8% of GDP. The political chaos and economic drag serve as a cautionary tale for any sovereignty decision — including those about on-premise digital infrastructure.

Anthropic’s new feature brings an always-on LLM to Slack. Beyond productivity, it’s a strategic move to capture organizational knowledge and workflows. AI-RADAR examines the data sovereignty challenges and what it means for those considering on-prem deployments.

German defense-tech startup Stark has raised €500 million in new funding, reaching a valuation of €3.5 billion. Backed by investors including Sequoia and the NATO Innovation Fund, over 80% of the capital will go directly into R&D and manufacturing to accelerate sovereign European defense capabilities — a move that underscores the continent’s increasing focus on autonomous strategic infrastructure.

The incident shows how a breach in a competitive intelligence vendor allowed hackers to access LastPass’s Salesforce environment, stealing personal information and support tickets. The encrypted vault remains secure, but the episode reignites the debate over supply chain security and access token management.

Early tests of the Linux 7.2 kernel on AMD EPYC Sorano reveal unexpected local network and socket performance improvements, alongside cache-aware scheduling. This signal could translate into greater efficiency and lower TCO for on-premise LLM inference workloads, strengthening data sovereignty.

A Breda perfume shop uses an algorithm to blend scents on the fly. Beyond the sensory appeal lies a critical question for those running proprietary AI: where does the model run? Protecting secret formulas pushes toward on-premise deployment, with all the hardware and control implications that entails.

Irish startup Ubotica has secured $11 million to speed up the commercial rollout of its orbital AI platform for maritime intelligence. The system processes data directly on satellites, slashing response times and improving threat detection across vast ocean expanses.

Public records reveal how U.S. libraries are avoiding LGBTQ+ displays and events for fear of backlash, practicing a creeping form of self-censorship. The firing of a Missouri librarian exposes political and religious pressures that echo tensions over content moderation in artificial intelligence.

The read-only EROFS filesystem introduces enhancements for handling large sparse AI datasets, reducing I/O overhead. A step forward for on-premise inference deployments, where every read cycle impacts TCO and latency.

Nvidia has unveiled a liquid cooling system that operates at unusually high temperatures, promising to cut electricity use and eliminate water consumption. The announcement highlights how thermal infrastructure is becoming a strategic lever for those evaluating on-premise deployment of high-density GPU clusters.

The Swedish startup bets on short video profiles and AI agents that screen candidates. Between social-media appeal and efficiency gains, questions arise about personal data handling and the role of on-premise deployment in regulated environments.

A leaked internal document from Madison Square Garden reveals how the company compiled tweets and public statements from activists opposed to its facial recognition system. Beyond ethical concerns, the breach underscores the risks of centralized surveillance infrastructure and reinforces the case for local, sovereign control over sensitive data.

Fwupd has released version 2.0.21, backporting fixes for more than 250 potential security issues discovered by AI-powered code analysis. The development underscores how automated scanning is reshaping software quality assurance—especially relevant for organizations managing firmware updates in on-premise environments.

Valve is developing SteamOS for broad PC release, teaming up with Nvidia to ensure compatibility. The company also hints at dual-boot capabilities down the road, pointing to a Linux-based OS that could appeal to users seeking locally optimized GPU environments — beyond just gaming.

Paris-based Tsuga, founded by former Datadog engineers, has raised a $35M Series A to build observability for the AI era. The startup aims to keep telemetry data inside the customer’s own cloud, pushing back against per-byte pricing as AI workloads cause a telemetry explosion, and prioritizing data sovereignty and cost control.