A London startup just raised $50 million to sell governments something they didn’t know they needed. Call it a sovereign AI control layer. It decides which AI systems get to touch sensitive data, and when. That’s one thread today. Meanwhile, the other is messier. A Bronx hospital replaced twelve nurses with AI software this month. In addition, the OECD says only 1% of workers actually have the advanced AI skills companies now want. Three stories, one theme. The infrastructure of AI oversight is being built right now, and most HR teams are watching from the sidelines.
Valarian Raises $50M to Build the Control Layer Governments Actually Trust
What Happened
Valarian is a London-based startup founded by Max Buchan and Josh McLaughlin. It closed a $50 million Series A led by NEA on July 13. That’s NEA’s first defense and dual-use investment in Europe, and it pushes Valarian’s total funding to $70 million. In addition, Lightbank, XTX Markets, Sequel, LitVC, and angels Gokul Rajaram and Nikesh Arora joined the round. (Source: Fortune)
Valarian doesn’t build another AI model. It builds the layer underneath. That’s software that governs how AI systems and other high-consequence applications access data, talk to each other, and operate. Specifically, two tracks exist: one for enterprise customers, one for defense. Think of it as a sovereign AI control layer. It sits between your AI tools and the sensitive systems they touch, enforcing rules at the infrastructure level instead of hoping every vendor behaves. (Source: NEA)
Why a Sovereign AI Control Layer Matters for HR
If your company pipes employee data into third-party AI tools, salary bands, medical leave records, you already have a version of this problem. You just don’t have Valarian’s version of the fix. Most HR tech stacks today grant broad API access to AI vendors and hope for the best. Valarian’s bet, backed now by visible support from the UK government, is that “hope for the best” stops being acceptable. That’s especially true once AI systems start acting autonomously instead of just answering questions.
Most mid-size companies already run five or six AI-enabled HR tools. An infrastructure-level control layer like Valarian’s is currently out of reach financially. That’s a venture-scale problem for now. But the underlying question is not out of reach. Who governs what your AI vendors can actually touch? That’s the sovereign AI control layer question, and it applies right now, at whatever scale you’re operating.
What to do: Ask your HR software vendors a blunt question this week. What data does your AI feature actually have write access to, and who audits it? If nobody at the vendor can answer quickly, that’s your answer too.
Montefiore Cuts Nursing Jobs, Hands Insurance Reviews to AI
Montefiore Medical Center eliminated 12 utilization-review nursing positions across its three Bronx campuses this month. The work of confirming that patient care is medically necessary now goes to AI software from Datavant. The New York State Nurses Association says the move violates AI-protection language the union won after a 41-day strike. In addition, it flags Datavant’s reported ties to Palantir as a data-exposure risk. Montefiore calls the union’s characterization “inaccurate and misleading.” (Source: Bronx Times)
So what? Utilization review is exactly the kind of judgment-heavy, rules-based work that AI tools for HR get pitched as ready to automate right now. Specifically, it’s the kind of work where a wrong call delays someone’s care. If you’re an HR or ops leader eyeing similar automation, pay attention. The Montefiore fight previews the same sovereign AI control layer question every workforce will eventually face: who decides when AI gets the final say, and who double-checks it.
OECD: Advanced AI Skills Still Belong to Just 1% of the Workforce
A new OECD policy paper, “Skills in the AI age,” has a stark finding. Firm-level AI adoption across member countries roughly tripled between 2021 and 2025, from about 7% to 20% of businesses. But workers with advanced AI skills, machine learning, data science, still make up only around 1% of the workforce. As a result, the OECD warns that without stronger training systems, this gap will widen inequality. Large firms will keep pulling away from small ones, and high-skill roles from routine ones. (Source: OECD, via Digital Watch)
So what? Your company’s AI adoption probably outpaced your hiring pipeline for people who can actually run the tools. In addition, that mismatch is why sovereign AI governance efforts like Valarian’s are competing for the same scarce talent pool everyone else wants. If you’re building out an HR function this year, treat AI skills training as infrastructure, not a nice-to-have workshop.
170,000 arXiv Papers Say Agentic AI Evaluation Is Now the Hot Field
AI Papers Academy analyzed 170,927 arXiv papers posted between January 2025 and June 26, 2026. The papers span computation and language, computer vision, machine learning, and AI research categories. It’s the first edition of their “AI Research Pulse” report, and it tracks share of papers rather than raw counts, because the field grew about 25% overall. Specifically, the findings flag agentic AI evaluation, AI-assisted scientific discovery, and autonomous research capability among the fastest-growing shares of research output in H1 2026. (Source: AI Papers Academy)
So what? Translation for non-researchers: the people publishing AI research are now spending more time asking “does this agent actually work” than “how do we build a bigger model.” That’s a healthy sign. It’s early evidence that the AI agents your vendors are pitching for HR workflows are getting more scrutiny before they ship, not less. That’s the same instinct behind Valarian’s sovereign AI control layer bet: trust needs verification, not just good intentions.
Quick Hits
- Monzo co-founder Tom Blomfield is taking a leave of absence from Y Combinator to join Anthropic’s compute team, working under co-founder Tom Brown. In particular, it’s another sign that frontier AI labs are recruiting operators, not just researchers, now that compute is the real bottleneck. (Source: TechFundingNews)
- Singapore’s dConstruct closed a $125 million Series A, one of the largest robotics raises in Southeast Asia to date. Specifically, the funding scales its autonomous construction robots, built to work in GPS-denied environments. (Source: JTC Singapore)
If today’s stories have you rethinking who actually has access to your HR data once AI enters the picture, that’s worth a real audit, not a guess. Asanify’s HR and payroll platform keeps AI-enabled workflows inside one governed system, instead of scattered across five vendor integrations you can’t fully see. Think of it as a smaller-scale, practical answer to the same sovereign AI control layer question Valarian is now selling to governments.
FAQ
What is a sovereign AI control layer?
A sovereign AI control layer is infrastructure that sits between AI systems and the sensitive data or operations they can access. It enforces rules about what an AI agent is allowed to touch, and when. Instead of building another AI model, companies like Valarian build this as a governance layer for governments and enterprises.
Will AI replace HR and healthcare administrative jobs?
Specific tasks are already being automated. For instance, Montefiore’s shift of insurance utilization review to AI software is one case. But wholesale replacement of licensed, judgment-heavy roles is contested territory. Unions and regulators are actively pushing back on where the line sits, so expect this to stay a fight rather than a settled fact.
How big is the AI skills gap right now?
According to the OECD, advanced AI skills like machine learning and data science are held by only about 1% of the workforce. That’s even though firm-level AI adoption roughly tripled between 2021 and 2025. As a result, the gap is a big reason AI-related roles stay open for months.
Not to be considered as tax, legal, financial or HR advice. Regulations change over time so please consult a lawyer, accountant or Labour Law expert for specific guidance.
