Aleph Alpha

• The mission is framed entirely around European sovereignty and EU regulatory compliance; data sovereignty is invoked but exclusively as a European jurisdictional and corporate concept, not as Indigenous Data Sovereignty.
• Public-facing copy centres government, defence, and industrial users as the relevant stakeholders.

• No acknowledgment of CARE Principles or Indigenous data governance.
• No engagement with embodied, relational, or oral knowledge traditions.
• 'Data sovereignty' is co-opted to mean European/enterprise control, displacing the term's Indigenous lineage without acknowledgment.

Score 1. Aleph Alpha's entire 'sovereignty' frame is European-state and enterprise-bounded. The lab uses 'data sovereignty' as a flagship term while making no reference to Indigenous peoples, CARE, or relational knowledge. The appropriation of the sovereignty vocabulary into a purely jurisdictional-commercial register, with zero Indigenous reference, warrants the floor score.

• There is unusual historical self-awareness about one axis: the lab explicitly positions itself against US/non-European cloud dependency, naming geopolitical infrastructure dependence as a problem ('uncompromisingly on European infrastructure').
• The 'AI Made in Germany. For Europe.' framing gestures at a specific geopolitical-industrial lineage (European digital autonomy).

• No acknowledgment of colonial data-extraction legacies or the history of the labour and resources (GPUs, minerals, annotation labour) underpinning the models.
• European sovereignty is presented as a clean, self-evident good with no reckoning with Europe's own extractive history.
• No transparency about regulatory constraints as historically contingent rather than given.

Score 3. The lab demonstrates partial historical consciousness — it names geopolitical infrastructure dependence and positions itself within a European digital-autonomy lineage. But this history is narrated triumphally; there is no humility about Europe's colonial inheritances, the material supply chain, or the contingency of the regulatory frame it leans on. Partial credit above the floor for naming a real historical-geopolitical dynamic.

• Multilingualism is a stated focus area, and the lab has a genuine technical history of European multilingual model development.
• 'For Europe' implies plural national/linguistic contexts within the EU rather than a single-language default.

• The homepage offers no concrete evidence of preserving culturally specific reasoning patterns — multilingual support reads as coverage, not cultural plurality.
• No mention of cultural scholars, humanities consultation, or non-Western epistemologies.
• 'Europe' functions as the outer horizon; cultures beyond the EU are absent.
• Western regulatory/categorical logic (compliance, regulation) is treated as the universal frame for trustworthiness.

Score 3. Multilingualism is real and central to the product, which lifts this above the floor. But on public copy it is framed as enterprise/jurisdictional coverage rather than as preservation of distinct cultural reasoning. The cultural horizon stops sharply at the EU border, and compliance logic is presented as the universal measure of trust.

• The lab leans on concrete, quantified impact claims (80,000 government users; 90% reduction in document search; 40% faster RFQ processing), giving an evidentiary surface.
• Evaluation tooling is mentioned ('evaluation tools ensure solutions integrate with organizational workflows').

• The quantified claims are unsourced, customer-context-free marketing metrics with no methodology, baseline, or independent verification.
• No independent audits of training data or bias disclosed.
• No third-party replication protocols.
• Closed/proprietary API; no open weights for external verification, despite the lab's earlier open-model history.
• No explicit known-limitation disclosures on the public page.

Score 3. There is an evidentiary posture — specific numbers and evaluation tooling — but it is the evidence of a sales sheet, not of science: no methodology, no third-party audit, no replication, closed weights. The presence of measurable claims keeps this just above the floor; the absence of verification or limitation disclosure caps it low.

• The register is uniformly instrumental: efficiency, compliance, reliability, business-critical outcomes.

• No acknowledgment of affective or intuitive dimensions.
• No space for ambiguity or poetic uncertainty — the copy is declarative and metric-driven.
• No recognition of emotional labour.
• Attention is framed entirely through efficiency (faster processing, reduced search times).

Score 2. The communication is purely efficiency- and compliance-coded. There is no gesture toward feel, ambiguity, or non-instrumental modes of attention. Marginally above floor only because the 'individualized' framing implies some attention to particular user context rather than pure throughput.

• European sovereignty is itself a future-modelling claim — a vision of who should control the AI infrastructure of the coming decade.
• Implicit governance frame via EU regulatory compliance.

• No engagement with labour displacement — indeed, efficiency/acceleration gains (faster processing, fewer search hours) imply displacement that is never named.
• No environmental or energy cost disclosure.
• Defence is named as a target sector with no discussion of the democratic governance of autonomous/agentic systems in that context.
• No inclusive or public deliberation; the future is shaped by and for governments, defence, and industry.

Score 3. Sovereignty is a substantive futures stance, and EU compliance is a real (if thin) governance commitment, lifting this above the floor. But the futures imagined are institutional and state/defence-centric; labour displacement is implied by the efficiency pitch yet unaddressed, environmental cost is silent, and the naming of defence without governance discussion is a notable omission for a future-modelling lens.

• The stakeholder set is explicitly governments, defence, semiconductor and automotive industry — institutional power, not marginalised constituencies.

• No participatory design with Global South developers (the horizon stops at Europe).
• No disability/accessibility commitments.
• No labour-representative engagement; workers appear only as the efficiency variable being reduced.
• No compensated community feedback channels.

Score 1. Every named stakeholder is an institution of state or industrial power. There is no table at which marginalised voices could sit — Global South, disabled users, and labour are wholly absent, and labour appears only as a cost to be cut. Floor score.

• The copy is uniformly solemn, confident, and self-consistent; no irony, self-questioning, or named contradiction.

• No willingness to name the obvious tension between 'responsibility' and a primary defence-sector customer base.
• No acknowledgment of the contradiction in branding 'data sovereignty' while running a closed, proprietary API.
• No space where the official narrative is tested by its opposite; the lab's seriousness is treated as exempt from audit.

Score 1. The communication is frictionless corporate solemnity. It smooths over the most obvious available contradictions (responsibility-branding + defence sales; sovereignty-branding + closed proprietary weights) rather than naming them. There is no inversion, irony, or self-audit. Floor score.