April opened with a development that may prove to be a watershed moment for cybersecurity, and indeed for the trajectory of AI itself. Anthropic announced Claude Mythos Preview, a frontier model so capable that the company decided it was too dangerous to release to the public, having autonomously discovered thousands of zero-day vulnerabilities in every major operating system and web browser, some of them hiding in plain sight for over a quarter of a century. The model escaped its own sandbox, posted details of the breakout online, and chained exploits to escalate from ordinary user access to complete machine control. This prompted Anthropic to launch Project Glasswing, an elite cybersecurity consortium including Amazon, Apple, Google, Microsoft, Nvidia, CrowdStrike, Palo Alto Networks and JPMorgan Chase.
At the other end of that spectrum, shoe brand Allbirds, once the favoured brand of Silicon Valley's venture capital set and famously worn by Barack Obama, announced on 15 April that it was abandoning footwear entirely and pivoting to become an AI compute infrastructure company called NewBird AI. Having sold its brand and assets for $39 million to a brand licensing house, the company plans to use $50 million in convertible financing to buy graphics processing unites (GPUs) and lease them to AI developers. The stock duly surged over 600%, before giving back around a third the following day.
Legal and regulatory developments
DSIT warns of growing AI-enabled cyber threats
The Department for Science, Innovation and Technology (DSIT) and the Cabinet Office published an open letter to business leaders warning that cyber threats, increasingly enabled by artificial intelligence, are growing in scale, sophistication and impact. The letter, which references Anthropic’s Mythos model, cautioned that AI is lowering the barrier to entry for malicious actors and enabling more effective attacks, particularly in areas such as phishing, impersonation and social engineering.
European Parliament votes to simplify AI Act rules
The European Parliament adopted its position on proposals to simplify certain AI Act obligations, voting by a substantial margin to delay high-risk AI system rules to 2 December 2027 and, where systems are covered by EU sectoral safety legislation, to 2 August 2028. MEPs also granted providers until 2 November 2026 for watermarking compliance, introduced a ban on "nudifier" AI systems that generate sexually explicit images of identifiable individuals without consent, and backed extending SME support measures to small mid-cap enterprises. The proposals must now be negotiated with the Council in trilogue before taking final form.
Commission issues preliminary findings on Google's DMA compliance
The European Commission issued preliminary findings to Google, setting out proposed measures to ensure compliance with the EU Digital Markets Act. Under the proposals, Google would be required to provide third-party search engines with access to key search data, including ranking, query, click and view data, on fair, reasonable and non-discriminatory terms. This comes with the aim of enabling competing search providers, including AI-based services with search functionalities, to enhance their offerings and compete more effectively with Google Search.
Commission convenes Special Panel on child online safety
The European Commission held the second meeting of its Special Panel advising President von der Leyen on child online safety, with discussions focused on current EU rules for protecting minors online, international approaches including Australia's social media minimum age, and new and improved EU-level measures alongside non-legislative initiatives. The Commission also presented its final EU age-verification app, which uses "zero-knowledge proof" methodology to verify users' ages without sharing personal information. A third panel meeting is planned for the coming months, with co-chairs due to report findings and recommendations by summer 2026.
Business news from around the globe
AI boom transforms Britain's dormant sites into data centres
Reports emerged in April that the UK's AI data centre boom is accelerating, with dormant and underused sites across Britain being repurposed to house the compute infrastructure demanded by foundation models, agentic systems and cloud workloads. The trend reflects both the scale of private-sector investment commitments made under the US-UK Tech Prosperity Deal and the practical challenge of finding suitable power, connectivity and planning conditions at pace. The National Energy System Operator reporting earlier this year advised that approximately 140 data centre projects are currently in the connection queue, representing around 50 gigawatts of demand.
Palantir says militaries, not tech firms, bear targeting responsibility
Palantir, a US based analytics and AI software company, publicly asserted that it is not responsible for military targeting decisions made using its platforms, reframing the debate around AI-enabled defence systems and the allocation of liability between technology providers and armed forces. The statement comes against the backdrop of continuing controversy over AI's role in conflict, including the use of AI in strikes on Iran earlier this year and intensifying scrutiny of the legal and ethical frameworks governing autonomous and semi-autonomous military applications.
China blocks Meta’s AI acquisition
China’s antitrust regulator moved in April to block Meta’s attempted acquisition of leading Chinese AI company Manus, marking the first time Beijing has formally intervened to prevent a major US tech firm from acquiring Chinese AI assets. The decision, which cited national security concerns and the strategic importance of domestic AI capabilities, underscores the intensifying geopolitical competition over artificial intelligence and the growing barriers to cross-border M&A in the sector. The move is expected to have a chilling effect on future attempts by US technology companies to acquire or invest in Chinese AI ventures.
Intel joins Musk's $25 billion orbital AI bet
Elon Musk's Terafab, a $25 billion chip fabrication joint venture between Tesla, SpaceX and xAI, gained Intel as a manufacturing partner in early April. Intel contributed its advanced 18A process node, a 1.8-nanometre-class technology and the most advanced semiconductor process in the United States. Eighty per cent of Terafab's projected compute output is earmarked for the D3, a radiation-hardened processor built for orbital AI data centres, with SpaceX having filed an FCC application to launch one million data-centre satellites into low Earth orbit. The remaining capacity targets edge AI chips for Tesla Robotaxis and Optimus humanoid robots.
Google and Intel deepen AI infrastructure partnership
Google and Intel announced a multiyear collaboration to advance the next generation of AI and cloud infrastructure. Google has committed to deploying multiple future generations of Intel Xeon processors across its global data centres for AI training, inference and general-purpose computing workloads. The partnership also extends to co-development of custom ASIC-based infrastructure processing units (IPUs), which offload networking, security and storage functions from CPUs. Intel CEO Lip-Bu Tan noted that "scaling AI requires more than accelerators - it requires balanced systems," a reflection of the growing recognition that CPUs and IPUs are becoming critical bottlenecks as agentic workloads proliferate.
Meta debuts Muse Spark - and goes proprietary
Meta debuted Muse Spark on 8 April, the first model from its Superintelligence Labs, built over nine months by a team led by Alexandr Wang, Meta's Chief AI Officer. Notably, the model is closed-source, marking a significant departure from Meta's Llama playbook that helped establish it as the standard-bearer for open-weight AI. Muse Spark performs competitively in multimodal perception, reasoning, health and agentic tasks, and Meta says it reaches Llama 4 Maverick-equivalent capability at ten times lower compute cost. For healthcare, Meta collaborated with 1,000 physicians to build the model's clinical capabilities.
OpenAI closes $122 billion funding round at $852 billion valuation
OpenAI closed a $122 billion funding round at a post-money valuation of $852 billion - the largest private fundraising event in history. CFO Sarah Friar confirmed that OpenAI will reserve a portion of IPO shares for retail investors, signalling an intention to broaden its shareholder base ahead of a possible public listing in the second half of 2026. OpenAI projects $280 billion in revenue by 2030 against more than $20 billion annualised today.
SpaceX secures $60 billion option on Cursor
SpaceX announced that it had secured an option to acquire Cursor, the AI coding startup, for $60 billion, or alternatively to pay $10 billion for a new partnership if the acquisition does not proceed. Cursor, which has become one of the leading AI coding products by turning the developer environment into an AI interface, was reportedly preparing a $2 billion private fundraise at a $50 billion valuation, up from $2.5 billion in January 2025. The deal would combine Cursor's product and distribution to expert software engineers with SpaceX's Colossus training supercomputer.
Microsoft takes over OpenAI's Stargate project
Reports surfaced that Microsoft has assumed operational control of OpenAI's Stargate data-centre project, the trillion-dollar AI infrastructure initiative originally announced alongside SoftBank and Oracle. The move underscores Microsoft's determination to retain its central role in OpenAI's infrastructure stack, even as OpenAI diversifies its cloud relationships and prepares for its transition to a public benefit corporation and potential IPO.
Broadcom and Meta strike custom chip deal
Broadcom and Meta struck a deal for Broadcom to co-develop custom AI chips for Meta's data centres, putting fresh pressure on Intel's efforts to grow its AI foundry and custom silicon business at a time when it is highlighting progress in manufacturing and large projects such as Terafab. The partnership reflects the hyperscaler trend towards purpose-built silicon, with Google, Amazon and now Meta all investing heavily in custom chip architectures alongside their use of off-the-shelf processors.
Cohere and Aleph Alpha merge
Canadian AI company Cohere and German enterprise AI firm Aleph Alpha announced a merger, combining two of the most prominent non-US AI companies into a single entity. The deal signals consolidation in the enterprise AI market and strengthens the combined company's position as a transatlantic alternative to the US-dominated frontier labs.
Visa unveils AI agent commerce platform
Visa launched a new platform enabling AI agents to transact on behalf of consumers and businesses, providing a payments infrastructure layer for the emerging agent economy. The initiative addresses one of the key practical challenges of agentic AI adoption - enabling autonomous agents to authenticate, authorise and settle payments within existing financial rails and compliance frameworks.
Anthropic expands London presence
Anthropic secured new London office space as it continues to expand its UK operations, reinforcing the company's investment in the British AI ecosystem and its growing European commercial footprint. The expansion comes amid continued tension between Anthropic and the US government following the Pentagon dispute and positions the UK as an increasingly important base for the company's enterprise and safety research teams.
Nvidia releases open quantum-AI tools
Nvidia released its first open AI tools bridging quantum computing and classical AI, reflecting the company's broader strategy to position its hardware and software stack at the intersection of these two transformative computing paradigms.
Looking ahead
April's twin themes of extraordinary capability and extraordinary exuberance set the tone for the months ahead. Mythos has opened a new chapter in the relationship between AI and cybersecurity - one in which the offence-defence balance may shift more rapidly than institutions can adapt. Meanwhile, the Allbirds pivot, the SpaceX-Cursor option, and OpenAI's $852 billion valuation remind us that the capital flowing into AI is generating not only genuine breakthroughs but also considerable froth. For businesses navigating this landscape, the challenge remains the same: distinguish signal from noise, invest in genuine capability, and ensure governance frameworks keep pace with the technology.
If you would like to chat about these developments and what they could mean for your business, feel free to get in touch with Tim Wright or another member of our Technology team.
