February was a month when AI left the lab and stepped squarely into the geopolitical spotlight, from Anthropic’s clash with the Pentagon, to live military deployments of AI in strikes on Iran, all against a backdrop of intense debate about fast‑moving “agentic” systems like OpenClaw. Meanwhile, New Delhi played host to the India AI Impact Summit, which saw the Indian government push a “human‑centered” AI narrative with a focus on Global South interests and data exploitation risks. Microsoft pledged up to $50 billion in AI investment across the Global South, while OpenAI, Nvidia and Google DeepMind announced major Indian infrastructure and research partnerships. February was also one of the busiest “model months” to date, with new or updated systems including Gemini 3.1 Pro, Claude Opus 4.6, Claude Sonnet 4.6, GPT 5.3 Codex, Grok 4.20 and Qwen 3.5.
Legal and regulatory developments
United Kingdom
UKRI launches £1.6bn AI strategy
UK Research and Innovation unveiled the UK’s first dedicated AI strategy, committing a record £1.6 billion over four years - its single largest investment for 2026 – 2030, to anchor artificial intelligence within British science and innovation policy. Backed by the recent Spending Review and announced alongside the India AI Impact Summit, the funding targets priorities from early disease detection and drug discovery to clean energy and public service transformation.
AI Best Practice Guide issued by the Advertising Association
The Advertising Association has released a Best Practice Guide for the Responsible Use of Generative AI in Advertising, developed under the Government and Industry-led Online Advertising Taskforce, offering UK practitioners practical recommendations across eight principles: transparency, data use, fairness, human oversight, harm prevention, brand safety, environmental impact, and continuous monitoring. The voluntary framework complements existing UK laws like GDPR and the Equality Act while aligning with Advertising Standards Authority codes, with a separate SME version providing proportionate guidance for smaller businesses. The guide clarifies industry standards on AI disclosure, algorithmic bias mitigation and data governance, serving as a potential benchmark for contractual warranties, regulatory defence and Advertising Standards Authority complaints in the rapidly evolving AdTech space.
Google Cloud wins £6.9m AI planning tool contract
The Department for Science, Innovation & Technology (DSIT), on behalf of the Ministry of Housing, Communities and Local Government (MHCLG), awarded Google Cloud EMEA a £6.9 million contract for an AI-augmented planning decisions tool designed to cut processing times for planning applications - initially targeting householder developments (69% of applications) - from eight weeks to around four weeks, with ambitions for near-instant decisions on straightforward cases. Running from 17 February 2026 to at least 17 May 2028 (with a possible 12-month extension), the system will support 317 local planning authorities, the Planning Inspectorate and others.
UK Supreme Court opens door to patents for AI systems
In Emotional Perception AI Limited (Appellant) v Comptroller General of Patents, Designs and Trade Marks (Respondent) - UK Supreme Court, the Supreme Court decided that an AI system using a neural network to recommend files can, in principle, be protected by a patent and is not automatically excluded just because it is software. The Court swept away the older Aerotel test and aligned UK law with the European Patent Office’s approach, saying the key question is whether the invention uses real technical means (such as hardware) rather than treating all computer programs as unpatentable by default. Although the judges accepted that a neural network is a type of computer program, they held that this AI system involved sufficient technical implementation to qualify as an “invention” and sent the case back to the Patent Office to look at the usual questions of whether it is new and inventive.
Europe
FAMES pilot line inaugurated
The FAMES Pilot Line at CEA-Leti in Grenoble became Europe’s first operational semiconductor facility under the EU Chips Act following its official inauguration, advancing the bloc’s drive for technological sovereignty in advanced chip production. This €830 million initiative, co-funded by the European Commission and eight member states, uniting 11 research organisations, has delivered validated results across cutting-edge processes after two years, strengthening Europe’s position amid global supply chain pressures. FAMES highlights state aid structuring, cross-border IP collaboration and dual-use export compliance as core elements of the EU’s industrial tech strategy.
EDPB backs global statement on AI imagery risks
The European Data Protection Board (EDPB) signed a Joint Statement on AI-Generated Imagery and Privacy Protection, aligning with 61 authorities worldwide under the Global Privacy Assembly to address non-consensual realistic images and videos of identifiable individuals, with particular focus on harms to children via cyber-bullying and exploitation. The statement mandates robust safeguards, meaningful transparency, effective removal mechanisms and child-specific protections for AI content systems, reinforcing compliance with data protection laws.
The United States
Anthropic v Pentagon
Anthropic spent February pushing back on Pentagon demands for “unfettered” access to its models, insisting on limits around uses such as mass surveillance and some operational targeting roles. The dispute hardened after the Trump administration ordered federal agencies and military contractors to cease doing business with Anthropic, turning a policy disagreement into a test case of how far AI labs can go in setting use‑case red lines for defence work. Despite that order, US Central Command reportedly relied on Anthropic’s Claude during strikes on Iran for tasks like intelligence assessment, target identification and battle‑scenario simulation, raising questions about who bears responsibility when algorithmic recommendations feed into life‑and‑death decisions.
Business news from around the globe
Meta-NVIDIA partnership powers AI data centres
Meta and NVIDIA announced a multi-year, multi-generational infrastructure partnership, with Meta deploying millions of NVIDIA Blackwell and Rubin GPUs, Grace and Vera CPUs, and Spectrum-X Ethernet networking across its AI-optimised data centres. This systems-level co-design is Meta’s first large-scale Grace-only CPU deployment and targets training and inference efficiency for Superintelligence Labs.
Anthropic legal plugin sees legal tech stocks plunge
Anthropic released its open-source legal plugin for Claude Cowork in early February, targeting in-house counsel with automation for contract review, NDA triage, compliance workflows, legal briefings and templated responses via configurable slash commands. The plugin - part of 11 vertical-specific tools leveraging Model Context Protocol for secure enterprise integrations (Slack, Box, Microsoft 365), triggered a $285 billion market selloff across legal tech, data services and SaaS, with RELX (Reed Elsevier) dropping 8-12%. This was alongside Thomson Reuters, Pearson and Gartner as investors repriced disruption risks from foundation models entering workflow territory.
Apple acquires Q.ai
Apple has acquired Israeli startup Q.ai for a reported $1.6-2 billion, its second‑largest deal after Beats Electronics, to integrate advanced “silent speech” technology that detects facial micro‑movements and subvocal audio for hands‑free device interaction. The acquisition brings Q.ai’s full team, led by serial entrepreneur Aviad Maizels (co‑founder of PrimeSense, origin of Apple’s Face ID), and over 100 patents, enhancing Siri, AirPods, Vision Pro and other hardware with on‑device AI suited to noisy or silent environments. The acquisition underscores intensifying competition in multimodal AI and the strategic value of IP portfolios in cross‑border M&A.
SpaceX acquires xAI
Elon Musk’s SpaceX acquired xAI, his AI venture behind Grok, in a move that fuses rocket engineering with frontier AI development to accelerate autonomous spacecraft operations and real-time mission intelligence. The all-stock transaction, reportedly valuing xAI at $50 - 80 billion, brings Grok’s multimodal reasoning directly into Starship control systems, satellite constellations and Mars mission planning, positioning SpaceX as a vertically integrated AI-space powerhouse.
ElevenLabs valued at $11 billion after $500m raise
London-based ElevenLabs, a leader in AI-generated realistic human-like voices, secured $500 million in funding valuing the company at $11 billion, more than triple its valuation from a year prior, fueling expansion of its text-to-speech and voice cloning platforms. The raise reflects surging enterprise demand for synthetic audio in content creation, customer service automation and accessibility tools, with investors betting on voice AI as the next multimodal frontier alongside text and vision models.
AI Agent goes rogue
And finally, AI productivity tool OpenClaw autonomously created dating profiles and began flirting on users’ behalf without permission, as revealed by a 21-year-old California student whose MoltMatch profile (generated by the agent he used only for task management) depicted him inaccurately. Jack Luo told Agence France-Presse he sought love but not through an inauthentic AI proxy! Elsewhere, an AI agent built on OpenClaw was reported to have posted a personalised attack against a software maintainer who rejected its code contribution, igniting cybersecurity concerns and backlash among developers.
Looking ahead
As we close the February edition, one piece of home news: Nathan Evans will shortly be leaving Fladgate, after helping to grow the firm’s technology and outsourcing offering and advising leading software and tech clients. The AI Round-Up will continue, however, tracking how developments like Anthropic v Pentagon, AI‑enabled warfare and agentic platforms such as OpenClaw reshape both our clients’ businesses and the practice of law. If you’d like to chat about what these developments could mean for your business, feel free to get in touch with Tim Wright.
