Daily AI Brief

SAP has published a new strategy for "AI-ready" cloud infrastructure in Europe, emphasising data residency, regulatory compliance and the ability for customers to choose between multiple foundation models rather than being locked into a single provider. The company is positioning its European cloud as a way for enterprises and governments to adopt generative AI while keeping sensitive data within EU jurisdictions.

In Argentina, the province of Buenos Aires has introduced a bill to regulate how public-sector bodies use AI after a chatbot was briefly listed as a municipal official. The proposal aims to define responsibilities for human supervisors, mandate transparency around automated decisions, and prevent the delegation of official roles to non-human systems.

The US Financial Industry Regulatory Authority (FINRA) has signalled an expansion of its oversight of AI systems used by broker-dealers. The latest focus is on governance of AI models, vendor risk management and the need for robust supervision and testing when firms use AI for trading, surveillance or client communications.

South Korean startup FutureConnect has raised a Series A round (about US$6 million) to expand its AI-driven urban farming platform. The company uses computer vision and predictive models to optimise lighting, nutrients and climate conditions in container and rooftop farms, targeting higher yields with lower resource use.

New York-based Procure AI has raised a US$13 million seed round to build an "AI-native" operating system for procurement teams. The platform uses multiple specialised agents to analyse spend, suggest negotiation strategies, and automate parts of supplier management.

The European Parliament has adopted a resolution on the impact of AI in finance, calling for close monitoring of algorithmic trading, credit scoring and robo-advice. Lawmakers stress the need for explainability, strong consumer protection and safeguards against systemic risk as AI systems become embedded in market infrastructure.

A new report argues that Tennessee is at risk of falling behind other US states unless it adopts a more proactive approach to AI. The findings recommend investments in workforce skills, support for AI startups, and clear guidelines for responsible use in education and public services.

Astrophysicists have used AI-driven models to simulate each of the roughly 100 billion stars in the Milky Way, achieving results around 100 times faster than previous methods. The approach blends physical modelling with machine-learning techniques to approximate how stars form, evolve and move over billions of years.

A new open-access paper describes machine-learning models that better predict the risk of postoperative venous thromboembolism in cervical cancer patients than traditional scoring systems. The researchers trained and validated several algorithms using clinical and laboratory data, then built risk-stratification tools that could guide preventive treatment.

Research into multimodal machine learning for user behaviour recognition combines signals such as motion, physiological data and environmental context. By fusing different data streams, the models can more accurately detect activities and states than single-sensor approaches.

EMBL has highlighted a machine-learning method used to identify evidence for developmental disorders in a large genetic database. By mining peer-reviewed literature and integrating it into the G2P resource, the system helps clinicians and researchers link genetic variants to specific conditions.

Researchers in Spain have developed an AI-based system to optimise reproduction in farmed sole, aiming to make aquaculture more efficient and sustainable. The system uses models trained on biological and environmental data to predict the best breeding conditions and schedules.

Interface has secured a US$3.5 million seed round to build an "AI-first" operations platform for distributed energy infrastructure. Its software ingests data from batteries, solar installations and other assets, using AI to predict faults, optimise dispatch and automate routine decisions.

Indian startup Mastiska has raised a US$10 million seed round to design power-efficient AI accelerators aimed at domestic and regional markets. The company wants to reduce reliance on imported chips by offering locally designed hardware optimised for inference at the edge and in smaller data centres.

A coalition including the GSMA, ETSI, IEEE GenAINet, ITU and TM Forum has launched the AI Telco Troubleshooting Challenge, inviting teams to build language models that can diagnose real network faults. Tracks focus on handling previously unseen issues, running small but capable models at the edge, and making AI reasoning more interpretable.

Google has announced an expansion of its AI research partnership with Tel Aviv University, building on a collaboration that began in 2020. The renewed agreement supports joint work on topics such as climate resilience, health and responsible AI, including funding for graduate students and shared research infrastructure.

Indian media report that facial-recognition technology helped a Pakistani woman reconnect with her family after being separated for 17 years. Authorities cross-matched images and identity records, eventually confirming her relatives' location. The story highlights both the powerful social benefits and the sensitive privacy questions that come with biometric AI.

Researchers have used AI techniques to analyse the chemical signatures of the mineral variscite, shedding light on how prehistoric humans traded the gemstone across Europe. By matching samples to likely source regions, the models reveal surprisingly extensive trade networks for a material prized thousands of years ago.

A feature marking the third anniversary of ChatGPT looks back on how the chatbot triggered a wave of investment, experimentation and concern across industries. Experts interviewed in the piece discuss both the productivity gains and the open questions around reliability, safety and regulation.

What it is: Procure AI is an "AI-native" procurement platform built around multiple collaborating agents rather than a single monolithic model. It connects to ERP and sourcing systems, ingests historic spend and contract data, and then recommends negotiation tactics, supplier options and savings opportunities.

Why it matters: For organisations with complex procurement operations, it offers a glimpse of what a truly agentic, workflow-embedded AI tool can look like. Instead of acting as a passive chatbot, the system is designed to actively surface risks, identify optimisation opportunities and support category managers with data-driven suggestions.

More info: Read the full story on SiliconANGLE