EU Tightens AI Rules as Model Labs Race to Build Compute

AI & Machine Learning

GitHub announced an update to Copilot’s interaction data policy that will, starting April 24, 2026, use inputs, outputs, code snippets, surrounding context and feedback from Copilot Free/Pro/Pro+ to train and improve its models unless individual users opt out; the change explicitly does not apply to Copilot Business and Enterprise customers. The post details what data will and won’t be used and explains opt-out mechanics while framing the change as a quality and security improvement for suggestions. The move tightens the feedback loop between live developer usage and model training, which could materially improve suggestion relevance but also raises renewed questions about developer trust and code provenance for those who do not opt out. How the community responds and whether regulators scrutinize the shift will shape wider norms for developer-tool telemetry and training data. Source: GitHub Blog Verified: True

China’s DeepSeek AI chatbot suffered its longest outage since its viral rise in early 2025, disrupting service for millions of users in late March and highlighting fragility in consumer-facing multimodal assistants. Reuters reporting indicates the extended downtime raised operational questions around capacity planning, content moderation backstops and the resilience of architectures built for rapid scale. For the sector, the downtime is a reminder that user trust and platform reliability are critical as chatbots take on broader consumer workloads, and that outages can have commercial and reputational costs for high-profile models. Companies operating mass-market assistants now face pressure to invest in redundancy, clearer SLAs and stronger incident communications to avoid similar disruptions. Source: Reuters Verified: True

Consumer Hardware

No major stories this sector today.

Cybersecurity

Reuters reported that the European Commission’s public Europa web platform was hit by a cyberattack on March 24 that disrupted cloud-hosted services and prompted immediate mitigation and investigation efforts. The incident temporarily affected externally hosted components and public-facing services, and the Commission said it took steps to contain the issue while assessing impact. The attack underscores ongoing pressure on high-visibility public cloud deployments and is likely to accelerate internal security reviews and external conversations about cloud vendor risk, redundancy and incident response for EU institutions. Public confidence in digital public infrastructure and the resilience of outsourced cloud services will be a key follow-on concern for policymakers and IT teams. Source: Reuters Verified: True

Medical-device maker Stryker confirmed a cybersecurity incident on March 11 that disrupted its Microsoft environment and global operations, and by late March said manufacturing and order-processing were “mostly restored” while investigations continued. Reuters coverage of the company update notes that some services remained impacted and that the outage prompted reviews across healthcare providers and regulators about supply-chain resilience and contingency planning for critical medical equipment. Attribution in some reports pointed to an Iran-linked hacktivist group, raising questions about geopolitical risk vectors affecting healthcare suppliers. The episode highlights how cyber incidents at a single OEM can ripple through hospital operations and regulatory oversight, reinforcing the need for stronger sectoral incident preparedness. Source: Reuters Verified: True

Enterprise Infrastructure

AWS published a March 31 service-availability update listing multiple features moving to maintenance or sunset and signaling end-of-support timelines for select legacy capabilities, along with migration guidance for customers. The notice affects a range of older and niche services and is a reminder that cloud vendors regularly prune functionality, requiring customers to track lifecycle announcements to avoid surprise disruptions or technical debt. Enterprises with long-running workloads must map dependencies, test migrations and factor vendor timelines into budget and architectural decisions to avoid operational headaches. The update is part of normal cloud lifecycle management but will force infrastructure teams to prioritize migrations in their 2026 roadmaps. Source: AWS What’s New Verified: True

Reuters reported that French AI startup Mistral raised about €780 million (around $830 million) in debt financing to scale data-center and infrastructure capacity for model training and inference. The financing underscores a push by European model labs to control more of their compute stack and reduce dependency on hyperscalers, aligning with broader sovereign and commercial ambitions for regional AI infrastructure. By locking in dedicated capacity, Mistral aims to accelerate model development and offer competitive hosting options to customers sensitive to data residency and vendor lock-in. The deal also signals to other European AI firms and policymakers that substantial capital is available for compute scale when strategic control of infrastructure is a priority. Source: Reuters Verified: True

CNBC reported a wave of quantum computing companies pursuing public listings and capital raises in late March, driven by investor and founder claims of progress on error correction, qubit scale and early commercialization pathways. Coverage highlights recent SPACs and IPOs (including firms such as Xanadu and Horizon Quantum) and frames the activity as an “inflection point” where research-led startups aim to become revenue-driven companies. Investors are betting on staged commercialization through cloud-accessed quantum services and software layers while cautioning that practical quantum advantage timelines remain uncertain. A successful shift to public markets would broaden capital access for hardware and software developers but also increase scrutiny on near-term milestones and commercialization roadmaps. Source: CNBC Verified: True

Reuters reported that BlackRock-managed funds injected roughly $57 million into Finnish quantum-hardware company IQM ahead of a planned U.S. IPO, reflecting continued investor appetite for quantum firms approaching public markets. The financing is positioned as preparatory capital to support scaling and market entry work prior to listing and signals confidence from institutional asset managers in select hardware plays. For IQM, the investment helps bridge the pre-IPO gap and provides validation ahead of a public debut, while illustrating how legacy financial firms are earmarking allocations to frontier technologies. The move further evidences a maturing private-to-public path for quantum startups and will likely influence valuation and competitive dynamics in the hardware ecosystem. Source: Reuters Verified: True

Policy & Regulation

The European Parliament adopted a package of amendments to the AI Act via the so-called “Digital Omnibus” in late March, imposing bans on AI systems that generate non-consensual sexual content (“nudification” tools), tightening rules for processing sensitive personal data, and adding registration requirements for certain high-risk systems. These changes move the AI Act closer to operational rules and enforcement, and they accelerate compliance timelines for vendors, platforms and service providers operating in the EU market. The amendments set the stage for trilogue negotiations with the Council and are likely to prompt product and policy teams to reassess data-handling practices, content-moderation tooling and risk-classification frameworks. For vendors, the outcome will drive design choices, transparency obligations and potential engineering investments to meet the tougher regulatory baseline in Europe. Source: European Parliament Verified: True