On Monday the Vatican published Pope Leo XIV’s first encyclical. On the platform with him was Christopher Olah, the atheist co-founder of Anthropic. The juxtaposition was the point. After two years of frontier AI being argued about as a technology question, the launch of Magnifica Humanitas reframed it as a moral one and put one frontier lab visibly inside the moral frame whilst its competitors stood outside. UK businesses watching the AI Bill debate should treat this as the most consequential rhetorical event of the year so far. The vocabulary your regulators, customers and employees will increasingly use was reset in public last week.
Strategic Reality: Anthropic’s presence at the encyclical launch is a positioning act, not a partnership. The lab is converting Trump’s recent public hostility — triggered by its refusal to license its models for warfare and mass surveillance — into a different kind of political capital. Being the AI company the Vatican will share a stage with is now part of Anthropic’s brand defence against the Musk-Trump bloc.
Why the encyclical matters more than a model release
Pope Leo XIV took his regnal name in deliberate echo of Leo XIII, whose 1891 encyclical Rerum Novarum did for the labour-capital question what the new document is attempting to do for AI. That is the right comparison. Rerum Novarum did not stop industrialisation, but it established the moral vocabulary in which industrial questions were argued for the following century: a just wage, the dignity of work, the responsibility of capital, the legitimacy of organised labour. The phrases became civic furniture.
Magnifica Humanitas is 42,000 words long and itemises the challenges posed by AI: displacement, autonomous warfare, surveillance, the question of who decides what gets automated and what does not. Its argument is that human dignity is the non-negotiable starting point. One speaker at the launch summarised the threat as a future in which most people are “user tools of an algorithmic order”. That phrase will travel.
| What the encyclical does | Why it matters for UK businesses |
|---|---|
| Establishes “humanity-first” as a moral primitive | Becomes the common framing across left, right, religious and secular UK audiences |
| Names specific harms (displacement, autonomy, surveillance) | Gives MPs and regulators a vocabulary that fits their existing portfolios |
| Endorses state regulation explicitly | Strengthens the political case for the UK AI Bill having teeth |
| Singles out reckless lab behaviour without naming Anthropic | Quietly partitions the frontier AI industry into good actors and bad actors |
| Quotes the duty to safeguard “the dignity of every person” | Becomes phraseology adopted by procurement frameworks and ESG reports |
Critical Context: The encyclical landed the same week the US president postponed an executive order that would have mandated safety reviews of new AI models. Read together, the two events sketch the political fault line UK AI policy will be drawn on: a deregulatory US administration aligned with parts of the industry, and a transnational humanity-first coalition that includes the Vatican, large parts of the EU, organised labour, and at least one frontier lab. UK regulators will be lobbied by both sides.
Why Anthropic went to the Vatican
The Guardian editorial called Anthropic’s presence “popewashing” — a fair charge, and one that businesses adopting Anthropic products should not duck. But the strategic logic is worth understanding on its own terms.
Anthropic has been excoriated by the US administration for refusing to allow its models to be used for warfare and mass surveillance. In US political terms, that puts the company on the wrong side of an administration that prizes loyalty and is willing to wield procurement, antitrust and immigration policy against firms that do not provide it. Sharing a stage with the Vatican does three things for Anthropic at once.
First, it converts a domestic liability — the Trump administration’s hostility — into an international asset. The same refusal that costs Anthropic in Washington buys it credibility in Brussels, Westminster, the Vatican, and the parts of the US federal procurement market governed by safety-first language.
Second, it puts a moral floor under Anthropic’s brand that its competitors will struggle to match. OpenAI has built a political-influence machine. Meta is positioned as the open-source champion. xAI is the loyalist frontier lab. Anthropic, distinct from all of them, is now the lab the Vatican will be seen with. That partitioning will outlast any individual product cycle.
Third, it gives Anthropic a cross-coalition vocabulary. The phrase “humanity-first AI” is durable in a way that “safety”, “alignment” or “responsible AI” is not. Safety and alignment are technical-sounding. Humanity-first reads the same in a parish newsletter, a Financial Times op-ed, and a Cabinet Office tender. Anthropic now owns the most useful AI phrase a non-technical audience has heard this year.
Strategic Insight: Frontier-lab positioning has bifurcated faster than most UK boards have noticed. The labs are no longer competing only on model capability. They are competing on which political and moral coalitions they can credibly stand inside. UK businesses choosing vendors are no longer making a pure technology decision — they are inheriting a political alignment.
The cost of this positioning to Anthropic is the popewashing charge itself, plus the awkward fact that Anthropic has been the most vocal frontier lab about Claude’s potentially “soulful” qualities, a claim that sits uneasily with Catholic doctrine on the unique status of human persons. The Vatican will have noticed. Its decision to share the platform anyway suggests it has made the same calculation Anthropic has: better to have the dialogue with the lab most likely to listen than to insist on perfect alignment with the labs unlikely to.
What it signals for the UK AI Bill debate
The UK AI Bill has spent two years circling the same set of questions: scope, sectoral versus horizontal regulation, the role of the AI Safety Institute, liability allocation, foundation model obligations, and how much to harmonise with the EU AI Act versus the lighter US approach. Until last week, the debate was conducted in technical language that excluded most of the public. The encyclical changes that on two fronts.
The first change is rhetorical. Any MP, peer or witness in the Bill’s committee stages can now invoke humanity-first as a non-partisan keyword. The phrase reaches Catholic, Anglican and Muslim communities; trade unions; the patient lobby; the disability lobby; education representatives; the creative industries. None of those constituencies have previously felt addressed by an AI policy debate. They will now.
The second change is positional. Frontier labs that have been arguing for light-touch regulation now have to argue against a moral frame, not just a technical one. The political cost of being seen to oppose humanity-first language has risen sharply. Expect to see vendor positioning papers in the next quarter that quietly drop “innovation-first” framing in favour of language compatible with the encyclical’s vocabulary.
| Stakeholder | What the new frame does to them | What they should do about it |
|---|---|---|
| UK regulators (Ofcom, ICO, FCA, MHRA) | Humanity-first becomes useable language in consultation papers and enforcement notices | Adopt the vocabulary in published guidance before vendors capture it |
| MPs and peers on the AI Bill | A cross-coalition frame that lets them speak to non-technical constituents | Use it explicitly; do not let it be ceded to a single lab |
| UK frontier-AI users (NHS, DWP, HMRC, councils) | Procurement language will tighten around dignity, autonomy and human oversight | Update vendor questionnaires before the Bill forces it |
| Businesses adopting AI | Customers and employees will read deployments through a humanity-first lens | Build governance that reads as humanity-first to non-technical audiences |
| Frontier labs other than Anthropic | Must develop competing moral vocabulary or accept partition | Watch for fast moves on faith-community, labour and patient engagement |
| Anti-regulation lobby | Loses access to the high-minded vocabulary it has used for two years | Expect a pivot to growth, sovereignty and competitiveness framings |
Competitive Reality: The UK AI Bill will pass into law in a public conversation that now includes the encyclical as a reference point. Vendors, advisers and lobbyists who continue to argue in purely technical or commercial terms will lose ground to those who can argue in both registers at once. This is a vocabulary contest as much as a substantive one — and the substantive provisions will follow whichever vocabulary wins.
How the humanity-first frame travels across UK policy contexts
The reason this matters operationally, rather than rhetorically, is that the same phrase is about to do work in policy areas that previously argued in different vocabularies. UK businesses should expect the following translations in the next twelve months.
NHS and adult social care. The phrase maps cleanly onto existing patient-dignity and Mental Capacity Act language. Expect NHS England procurement to include humanity-first criteria in AI tendering before the AI Bill is enacted. Suppliers will be asked to demonstrate that their tools augment clinical judgement rather than replace it, and that patient autonomy survives algorithmic recommendation.
Employment law and HR practice. The encyclical’s emphasis on the dignity of work travels straight into ACAS guidance territory and the ongoing debate about algorithmic management. Expect the next iteration of the Employment Rights Bill, or the consultation that follows it, to test whether humanity-first language belongs in statutory codes governing AI-driven performance management, scheduling and dismissal.
Public sector AI deployment. The Cabinet Office’s existing Algorithmic Transparency Recording Standard already gestures in this direction. The encyclical will accelerate departmental adoption and raise the political cost of high-profile failures. Expect DWP, HMRC and the Home Office to face sharper parliamentary scrutiny of AI-driven decisions affecting individuals.
Financial services. The FCA’s Consumer Duty framework already implies a humanity-first reading of AI used in pricing, underwriting and complaints handling. Expect supervisory letters within the next two quarters that make the connection explicit, and a tightening of expectations on vulnerable-customer handling where AI is in the loop.
Education. The Department for Education’s existing position on generative AI in schools is more permissive than the encyclical frame would suggest. Expect a tightening, particularly around assessment, safeguarding and the use of AI tools by children with special educational needs.
Creative industries. The encyclical’s framing of dignity travels into the ongoing fight over training-data scraping, voice and likeness rights, and the UK’s data-mining exception. The argument that artists should have meaningful control over how their work trains AI now has a moral spine the industry side will struggle to match without conceding ground.
Implementation Note: A useful test for any UK business adopting AI in the next six months is to draft a one-page humanity-first statement for each deployment — what the system does, who it affects, how human oversight is preserved, what recourse the affected person has. Treat this as a working document for procurement, employee comms and regulator engagement, not a polished marketing artefact. The discipline of writing it will surface gaps before a regulator or a journalist does.
Hidden challenges UK leaders should plan around
Four non-obvious risks the surface debate will not flag.
Humanity-first as marketing versus humanity-first as substance. The gap between language and practice is the risk that lands hardest on UK adopters. A vendor describing its product in humanity-first terms whilst its model card lists no meaningful human oversight provisions is a procurement red flag, not a positive signal. Mitigation: insist that humanity-first claims are accompanied by concrete operational specifics — who oversees, who can override, who is liable, what is logged.
The frame is morally durable but technically vague. “Humanity-first” does not, on its own, settle any of the difficult AI questions: when is automation appropriate, how should liability be allocated, what counts as meaningful human oversight, how is consent to algorithmic processing constructed. The phrase can paper over genuine disagreement and produce policy that everybody agrees with whilst nobody is clear what they have agreed to. Mitigation: when humanity-first language appears in a tender, contract or policy document, ask for the operational definition of each term it modifies.
Anthropic’s positioning may flip if US politics changes. A future US administration that values frontier-AI alignment less hostile-ly than the current one would remove the political incentive for Anthropic’s Vatican posture. The lab’s competitors will exploit any sign that the moral commitment is positional rather than structural. UK businesses that have publicly aligned with Anthropic’s humanity-first framing should plan for the possibility that the vendor’s stance moderates. Mitigation: tie your governance to the principles, not to the vendor that currently champions them.
UK procurement vocabulary will internalise the frame faster than UK statute. Procurement guidance moves on quarters; primary legislation moves on years. Expect humanity-first language to appear in Cabinet Office, NHS England and FCA documents long before it appears in the AI Bill on the statute book. Businesses that wait for the law to settle will already have been judged against the procurement frame for two or three buying cycles. Mitigation: read procurement guidance as the leading indicator, not the trailing one.
Hidden Cost: The popewashing critique applies most sharply to Anthropic, but it will be redirected at any UK business that adopts Anthropic products and adopts the humanity-first language without operational substance behind it. The reputational risk is shared. Customers and employees who learn to read humanity-first language as branding rather than commitment will eventually punish both vendor and adopter.
Strategic takeaway
The encyclical is not a theological event for UK businesses. It is a vocabulary event, and vocabulary shapes regulation more reliably than technical analysis does. The labs and the lobbyists understand this. The UK businesses adopting AI products inherit the language their vendors choose to speak.
The Vatican–Anthropic moment is the clearest signal yet that frontier AI is now contested ground in a way that runs across the political spectrum. The contest will not be won by whichever model benchmarks highest. It will be won by whichever moral and political coalition can credibly claim that the AI being deployed serves human flourishing rather than degrading it. UK businesses are participants in that contest whether they recognise it or not.
Take Action: The UK AI Bill consultation cycle continues over the next eighteen months. The window in which non-vendor voices can shape the vocabulary the Bill speaks is open now and closing. Brief written submissions matter. So does the language your business chooses for its own AI deployments — every contract, employee notice and customer-facing disclosure is a small bid in the vocabulary contest.
Three success factors for UK leaders engaging with this:
- Speak both registers. Technical, commercial language earns credibility with engineers and procurement. Humanity-first language earns credibility with regulators, MPs and the public. Strategic AI literacy now means moving fluently between them.
- Operationalise the frame. Humanity-first as a slogan is cheap. Humanity-first as a set of operational commitments — oversight, override, audit, recourse — is the version that survives scrutiny.
- Read the political layer, not just the product layer. Vendor announcements are the front of house. Where vendors are choosing to stand, and which coalitions they are joining, tells you more about the regulatory environment you will operate in than any benchmark.
Next steps for AI-adopting businesses:
- Draft a one-page humanity-first statement for each material AI deployment in your organisation
- Update vendor questionnaires to require operational specifics behind humanity-first claims
- Engage with at least one UK AI Bill consultation in 2026 — even a brief written response counts
- Audit existing AI vendor contracts for human-oversight provisions, override rights and recourse mechanisms
- Identify two or three independent AI policy intelligence sources you can monitor monthly, including faith-community, labour and patient-advocacy voices
Source and attribution
This analysis builds on The Guardian editorial “The Guardian view on the Pope and Claude: Leo XIV’s encyclical on AI is right to put humanity first” (25 May 2026): https://www.theguardian.com/commentisfree/2026/may/25/the-guardian-view-on-the-pope-and-claude-encyclical-on-ai-is-right-to-put-humanity-first. The interpretation, UK framing and strategic implications are Resultsense’s own. Our insights archive tracks ongoing developments in UK AI policy and frontier-lab positioning.