Research topic · Papers 1.0 · Demonstration 0.1.0

Europe’s answer to Palantir

A globally collaborative model for industrial AI

Author
Affiliation
Dr. Carsten Stöcker — Spherity GmbH
Published
Updated
Research cut-off
· Paper research cutoff; papers and fictional model dated 1 November 2026
This version
https://spherity.github.io/spherity-research/europes-answer-to-palantir.html
Latest version
https://spherity.github.io/spherity-research/europes-answer-to-palantir.html
Browse this paper

In brief

What does this research establish?

Palantir shows the power of connecting data, business meaning, logic and governed action through an operational ontology. Europe’s strategic opportunity is to make comparable cooperation work across independently governed companies and platforms. This research proposes access-controlled, verifiable evidence graphs built on open international standards and industry-agreed semantics. Participants prepare their own information, exchange permitted evidence and retain authority over consequential actions. Europe can lead this effort with partners in the US, Asia-Pacific and other regions. Existing dataspaces are a first step: fast-track modernisation and demanding operational tests are still needed for AI-first use.

Key takeaways

  • Preserve operational depth. Useful industrial AI needs interpreted relationships, evidence and controlled actions, not just more files.
  • Make cooperation portable. Shared meaning, discovery and trust should connect authorized participants across providers, without making one intermediary indispensable.
  • Prove the next step. Test performance, authorization, evidence freshness and provider substitution together. The papers propose this programme; they do not report a finished global infrastructure.

Two papers, one cooperation agenda

Executive Brief

Read the Executive Brief (PDF, 19 pages)

A European Answer to Palantir
A globally collaborative counter model for industrial AI

Version 1.0 · 1 November 2026 · 30 references · CC BY 4.0

Academic paper

Read the academic paper (PDF, 45 pages)

A Global Cooperation Model for Sovereign Industrial AI
European dataspaces and open standards for verifiable evidence across platforms

Version 1.0 · 1 November 2026 · 101 references · CC BY 4.0

Explore the robot evidence graph · Academic abstract and citation record

Complete first page of A European Answer to Palantir, version 1.0
A European Answer to Palantir · 19 pages · Dr. Carsten Stöcker, Spherity GmbH
Complete first page of A Global Cooperation Model for Sovereign Industrial AI, version 1.0
A Global Cooperation Model for Sovereign Industrial AI · 45 pages · Dr. Carsten Stöcker, Spherity GmbH

What is Europe’s answer to Palantir?

Palantir shows the power of connecting data, business meaning, logic and governed action through an operational ontology. Europe’s strategic opportunity is to make comparable cooperation work across independently governed companies and platforms. This research proposes access-controlled, verifiable evidence graphs built on open international standards and industry-agreed semantics. Participants prepare their own information, exchange permitted evidence and retain authority over consequential actions. Europe can lead this effort with partners in the US, Asia-Pacific and other regions. Existing dataspaces are a first step: fast-track modernisation and demanding operational tests are still needed for AI-first use.

Sovereignty means the practical capacity to choose partners, govern permitted uses and preserve essential operations. Palantir and other capable providers can participate. Shared vocabulary, discovery and trust operators can also become concentration points: credible substitution and continuity need testing.

Two architecture patterns

P1 is the author’s functional synthesis of documented Palantir capabilities (baseline 0.5). P2 proposes participant-controlled cooperation (baseline 0.4). Neither was benchmarked in this research.

Question Palantir-centred pattern, documented reference Participant-controlled pattern, proposed extension
What creates value? Operational ontology connects data, objects, relationships, logic, permissions and actions. Comparable local capabilities cooperate across independently governed implementations.
Where are files processed? Configured source access and preparation feed the operational model. Each participant prepares its own sources before authorized exchange; no compulsory shared ingestion centre.
Can several companies participate? Yes. Palantir documents multi-organization ecosystems and can use industry vocabularies. Yes. Shared exchange contracts aim to work across independently chosen implementations.
How is meaning shared? Configured platform models, mappings and operational interfaces. Maintained industry-agreed vocabularies, mappings and profiles across providers.
Who controls action? Configured platform and organizational permission boundaries. Each responsible participant retains local decision and action authority.
What is the strategic test? Evaluate useful operational integration and its lifecycle and exit costs. Demonstrate preserved meaning, evidence and behavior when a replaceable provider changes.
What has this research proved? It synthesizes documented architecture; no authenticated deployment benchmark was run. It defines a design and test agenda; no composed production federation or superiority result is reported.

“Knowledge Graph Builder” is an author-defined functional grouping, not a separately marketed Palantir product. See the detailed P1 and P2 diagrams.

Both patterns ingest varied data formats; each ecosystem participant processes its own files before controlled exchange and composition of a permitted cross-company graph view.
Each participant ingests, checks and links its own records. Its gateway exchanges authorized requests, evidence and results. An authorized requester composes selected contributions into a cross-company graph view, preserving sources and use conditions. Shared meaning, discovery and trust serve all participants. This is a task-specific view, not a pooled database. Local authority controls actions. The legend identifies formats, not trust. Palantir supports multiple organizations; the comparison concerns governance across implementations. Proposed architecture, not measured performance.

From records to meaning, evidence and governed decisions

Well-maintained graphs can help retrieve the relevant facts, preserve provenance, reuse explicit checks and show which decisions depend on changing evidence. This can reduce repeated document extraction, irrelevant prompt context and unsupported answers. Whether it reduces tokens, latency, cost or risk must be measured on the actual task, including graph maintenance, distributed retrieval and verification overhead. Graphs are not required for every AI task and do not eliminate hallucinations.

Open-world semantics allows knowledge to be incomplete: an unavailable report is not proof that it never existed or that the product failed. A bounded operational rule can still require accepted evidence before authorizing action. Open semantics does not mean public data or unrestricted access.

What changes when one test report is withdrawn?

Select a record in this fictional humanoid robot example to follow its relationships across a supplier, laboratory, manufacturer and operator. The graph connects components, configuration, test assertions, original reports, scoped authority and operational events. It makes the evidence behind a decision inspectable. It also shows where information is missing or a fresh review is required.

Fictional educational demonstration · Model 0.1.0 · 1 November 2026. All records are pre-disclosed. Filters and scenario controls illustrate the architecture; they do not enforce access control, verify signatures, run an AI model or establish regulatory compliance.

The interactive view needs JavaScript. The complete walkthrough, figure and records below remain available.

Follow the evidence: a static walkthrough

  1. Start with Configuration R3 (M02) of fictional humanoid H-01 (M01), performing nonmedical picking task TP-PICK (M03).
  2. Follow test assertion TA42 (L01) to its original report TR42 (L02), test method and laboratory scope. A report, an assertion and a release decision are different records.
  3. Follow release claim RC3 (M04) and the illustrative acceptance rule AR1 (M05). The rule requires current test support for the proposed decision.
  4. In the default Withdrawn scenario, TR42 is no longer accepted as a current premise under AR1. RC3 becomes HOLD, and dependent assessment/work-order records require review. Preserve the historical report and unrelated records. HOLD does not establish that the robot is unsafe or legally noncompliant.
  5. Switching to Available for review leads to REVIEW, not an automatic approval. The responsible actor still checks scope, conditions, permission and other evidence.
  6. Change the assessment lens to see different questions. A record can contribute to several assessments without creating one universal certificate or verdict.

The seven lenses are ESPR/DPP, functional safety, EU AI Act, MDR medical branch, Machinery Regulation, Cyber Resilience Act and export controls. These are not seven universally applicable duties. Product scope, intended use, dutyholder, jurisdiction and dates need separate assessment. This is not an exhaustive legal inventory.

Fictional robot configuration links supplier, laboratory, manufacturer and operator records; withdrawal of report TR42 highlights dependent claims requiring reassessment.
A synthetic humanoid evidence graph connects configuration R3 to components, a test assertion, its original report, scoped authority and operational events across four independently governed participants. Shared semantic, discovery and trust roles supply authorised metadata to all participants; depicted cross-boundary relationships are logical references, not unrestricted data flows. The highlighted trace shows a withdrawal of report TR42: under the illustrative rule AR1, release claim RC3 becomes HOLD and dependent assessments require review. This preserves historical records and does not establish that the robot is unsafe or legally noncompliant. All identifiers, sources, credentials, events and outcomes are fictional; no signatures or engineering results have been validated.

Download the authoritative graph model (JSON, 0.1.0). All four participants connect to Shared Meaning (META1), Discovery (META2) and Trust/status (META3). Cross-boundary relationships are logical references, not unrestricted data flows. The teaching dependency projection is separate from the 20 typed relationships; it does not prove causation or legal invalidation.

All 22 fictional records and their evidence conditions
S01 · Actuator A17 · Component supplier
type
entity
claim
Physical component, serial A17. Claims must identify the component and version actually installed.
limit
Asset identity; not a safety certificate.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Component supplier
S02 · Firmware FW3.2 · Component supplier
type
entity
claim
Versioned software installed on actuator A17. An update requires a fresh assessment of which evidence still applies.
limit
Version identity; no automatic transfer of a previous test result.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Component supplier
S03 · Software inventory SB04 · Component supplier
type
document
claim
Machine-readable software bill of materials for FW3.2, including dependency identifiers and versions.
limit
Inventory completeness and vulnerability relevance still require checks.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Component supplier
S04 · Material declaration MD2 · Component supplier
type
document
claim
Supplier assertion about the identified component or batch, retained with its source and issuer.
limit
A signed claim is attributable, not automatically true.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Component supplier
L01 · Test-result assertion TA42 · Independent laboratory
type
assertion
claim
The tested configuration met the selected criterion under method TM1 and task conditions TP-PICK. This is a fictional, bounded result.
limit
Subject, method, conditions and current admissibility must match.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Independent laboratory
L02 · Original test report TR42 · Independent laboratory
type
document
claim
Original fictional PDF from which TA42 was extracted. Its source identity and history remain available after withdrawal.
limit
Demonstration only; no real report or signature is supplied.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Independent laboratory
L03 · Test-method profile TM1 · Independent laboratory
type
policy
claim
Defines measurement units, test procedure, operating conditions and acceptance criterion for the example.
limit
A method profile is not accreditation or product certification.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Independent laboratory
L04 · Laboratory scope certificate LC7 · Independent laboratory
type
certificate
claim
Illustrative credential identifying a laboratory and the scope for which the verifier accepts its evidence.
limit
Certificate authenticity, scope, expiry and issuer acceptance are separate checks.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Independent laboratory
L05 · Report-status event RS9 · Independent laboratory
type
event
claim
In the withdrawn branch, a later event withdraws TR42 from current reliance. The available branch is an alternative scenario before that withdrawal; it does not erase the historical report.
limit
The receiving participant must obtain current status within its declared freshness bound. No safety verdict follows from withdrawal alone.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Independent laboratory
M01 · Humanoid robot H-01 · Robot manufacturer
type
entity
claim
Fictional industrial humanoid used to explain evidence reuse. Its intended task is non-medical material handling.
limit
No real manufacturer, product performance or endorsement is implied.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Robot manufacturer
M02 · Robot configuration R3 · Robot manufacturer
type
entity
claim
The specific robot, actuator, firmware and task configuration being assessed.
limit
A graph connection does not transfer a certificate to a new configuration.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Robot manufacturer
M03 · Task and operating conditions TP-PICK · Robot manufacturer
type
policy
claim
A bounded industrial picking task with specified workspace, payload, speed and supervision. Values are deliberately unspecified.
limit
New intended uses or operating conditions require applicability review.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Robot manufacturer
M04 · Bounded release claim RC3 · Robot manufacturer
type
assertion
claim
A proposed assurance claim for the selected configuration and task, supported by explicit accepted premises.
limit
This is neither a universal compliance certificate nor permission to operate.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Robot manufacturer
M05 · Manufacturer approval rule AR1 · Robot manufacturer
type
policy
claim
The example rule requires a current, applicable test result. Withdrawal of its only accepted report produces HOLD pending reassessment.
limit
Author-created teaching rule; not a claim about a legal approval procedure.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Robot manufacturer
M06 · AI evidence-assistant service AS1 · Robot manufacturer
type
entity
claim
A versioned virtual service that can retrieve permitted evidence, follow links and prepare source-linked explanations.
limit
The visualization runs deterministic rules; no live AI model or action authority is attached.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Robot manufacturer
O01 · Installed configuration IC3 · Industrial operator
type
entity
claim
The operator records which physical robot and software configuration are actually installed.
limit
Installation identity and operational context must match the assessed configuration.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Industrial operator
O02 · Maintenance event ME5 · Industrial operator
type
event
claim
A lifecycle event records work on the installed configuration and may trigger targeted reassessment.
limit
An event can invalidate assumptions without invalidating every certificate.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Industrial operator
O03 · Supply-chain delivery event DE7 · Industrial operator
type
event
claim
Receipt of the identified component connects its supplier provenance to the operator lifecycle.
limit
Custody, origin, export classification and permitted end use remain distinct questions.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Industrial operator
O04 · Work order WO9 · Industrial operator
type
document
claim
A request to use the installed robot for the specified task. It needs local authorization as well as accepted evidence.
limit
The graph cannot grant legal or operational authority by itself.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Industrial operator
META1 · Shared semantic profile SP1 · Multiple independently governed metadata services
type
policy
claim
Versioned industry definitions, identifiers, units and mappings make assertions interpretable across participants.
limit
Open standards need maintained profiles, mappings and conformance tests.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Multiple independently governed metadata services
META2 · Permitted evidence catalogue CAT1 · Multiple independently governed metadata services
type
document
claim
Authorized discovery of relevant evidence and services. Restricted existence and metadata must also be protected.
limit
The visible demo contains only pre-disclosed fictional metadata.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Multiple independently governed metadata services
META3 · Issuer, scope and status bundle TB1 · Multiple independently governed metadata services
type
certificate
claim
Accepted issuer references, credential scopes and current status information support local verification.
limit
Signatures, authority, freshness, applicability and permission are independent dimensions.
source_kind
synthetic teaching record
version
0.1
signature
Not cryptographically checked in this demonstration
access
Pre-disclosed example; a real query requires task-scoped authorization
status
Illustrative
owner
Multiple independently governed metadata services
All 20 typed relationships
Source recordRelationshipTarget record
M01has configurationM02
M02includesS01
M02usesS02
S03describesS02
S04concernsS01
L01testsM02
L01recorded inL02
L01used methodL03
L02issued underL04
L05updates status ofL02
M04assessesM02
M04for taskM03
M04appliesM05
M04requiresL01
M06explainsM04
O01instantiatesM02
O02concernsO01
O03deliversS01
O04requests use ofO01
O04requiresM04
Seven separate assessment lenses and scope limits

ESPR / DPP

Can product identity, materials and lifecycle claims be traced? ESPR DPP duties depend on applicable product measures and dates; a humanoid does not automatically require an ESPR DPP.

Functional safety

Does accepted evidence support this configuration in this task? Functional safety is a discipline and assurance task. A single test result does not establish system safety.

EU AI Act

What is the AI function, intended use and applicable assurance route? Classification, duty holder and sectoral route require assessment; a robot or AI assistant is not automatically high-risk.

MDR · medical branch

Would a medical intended purpose change the required evidence? The base case is non-medical. A medical variant would require its own classification, clinical evidence and conformity route; these are not supplied here.

Machinery Regulation

Which risk assessment, technical documentation and conformity route apply? Check the placing-on-market date and applicable machinery regime. The figure does not grant conformity or CE marking.

Cyber Resilience Act

Which software, vulnerabilities, support duties and incidents matter? Determine product scope, exclusions and staged duties. An SBOM is useful evidence, not a complete CRA assurance case.

Export controls

What classification, destination, end user and end use must be checked? No automatic licence requirement follows from the word robot. List entries, catch-all provisions and other relevant restrictions need case-specific review.

Fast-track modernisation for AI-first workloads

Europe’s dataspace work is a foundation for coordination and experimentation. It is not evidence that current implementations already satisfy IPCEI-AI or frontier-AI throughput, latency and volume requirements. The goal is fast-track modernisation without relaxing access control, verifiability or participant authority. Retain components that pass defined tests.

The meta-platform argument is participant-controlled cooperation across platforms. It does not imply a proven exponential network-value law, guaranteed GDP uplift or automatic efficiency gains from decentralization. Shared vocabulary, discovery or trust operators can themselves become concentration points; credible substitution and continuity need testing.

A Participant Exchange Gateway manages permitted exchanges. A dataspace connector built with Eclipse Dataspace Components is one implementation option for its exchange functions, complemented by identity, evidence verification and action authorization. Shared Meaning provides semantic and vocabulary-hub functions. These are logical shared roles, potentially with several providers. Every participant connects to Shared Meaning, Discovery and Trust/status.

ODRL is a policy-expression standard used in dataspace implementations, and EDC provides a reference implementation of policy evaluation. Neither ODRL nor a connector alone is a universal, complete enforcement or legal-recognition mechanism. Shared vocabularies, catalogues, resolvers and trust metadata also need authenticated publication, integrity checks, scoped access and explicit freshness bounds. Cached or ledger-backed information does not remove freshness or accepted-authority decisions.

Europe as a global convener

Europe can turn its experience in industrial dataspaces into a globally collaborative offer. The next step is shared implementation profiles, joint testing and accountable recognition arrangements with international partners. W3C work, UNECE traceability initiatives, US innovation and Asia-Pacific programmes provide relevant cooperation channels. Technical interoperability, accepted issuer authority and legal recognition are separate achievements; a technical connection does not establish all three.

Figure library: 12 originals, 14 paper placements

All 12 figures are available as SVG, vector PDF and PNG. The downloads include descriptive and licensing metadata; the artwork is unchanged. The two robot figures appear in both papers and have one public copy each. Open a full-size diagram to inspect small labels.

Seven distinct assessment lenses reuse selected evidence through permitted retrieval, verification, interpretation, explanation and locally authorized action.
Selected evidence can be reused across seven separate assessment views: ESPR/DPP, functional safety, the EU AI Act, MDR where medical scope applies, the Machinery Regulation, CRA and export controls. Each view requires its own applicability analysis, accepted evidence, authority and decision rule; the arrows do not assert that every record is relevant to every regime. The proposed workflow separates permitted retrieval, verification, interpretation, AI-assisted explanation and locally authorised action. A withdrawn source triggers review only where it supports a dependency. Reductions in repeated work, response time or cost and improvements in explanation or portability are hypotheses to measure, not observed outcomes or automatic compliance.
Platform-governed ontology capability compared with independently governed participants connected through shared meaning, discovery, trust and scoped exchange.
Conceptual comparison of ontology capability within a platform governance scope and across independently governed participants. Both require an ontology model, graph construction and an operational runtime; local ingestion and control loops are abbreviated. The proposed ecosystem connects independently governed participants through shared meaning, discovery and trust, scoped exchange and cross-company AI. Palantir supports multiple organizations; the left-hand focus does not assert a single-company limit or universal closed-world reasoning. On the right, open-world treatment of incomplete knowledge coexists with local authorization and bounded decision gates. This is a proposed extension, not a measured superiority result.
Twelve functional responsibilities connect source preparation, ontology model, graph builder, runtime, AI and controlled actions within a logical platform governance scope.
P1 integrates the twelve common responsibilities within a logical platform governance scope. The illustration permits private and shared organizational scopes and configured external services; it does not impose a single-company or single-database limit. The Knowledge Graph Builder is an author-defined functional grouping that overlaps the Ontology Engine write path. Principal data, configuration, request and action paths are shown; Appendix A records the detailed flow groups. Author synthesis from the documented P1 baseline.
Manufacturer, supplier and laboratory each connect local ontology capabilities to shared meaning, discovery and trust, with scoped exchanges through participant gateways.
P2 repeats local model, builder and runtime responsibilities across independently governed participants. The manufacturer, supplier and laboratory each connect to all three shared roles: Shared Meaning, Discovery and Trust. The purple metadata connections are distinct from scoped business exchanges through Participant Exchange Gateways. A dataspace connector built with Eclipse Dataspace Components (EDC), complemented by identity and evidence services, is one gateway implementation example. Shared roles may have several operators; they imply neither a mandatory central broker nor unrestricted access. This is a proposed composition, not an executed integration. Source: Eclipse Dataspace Components (linked in Sources and method).
Illustrative valve records connect to a measurement, source and acceptance rule; evidence and policy checks support an accountable decision rather than automatic approval.
A proposed valve-release example shows how heterogeneous records become decision-relevant evidence. The graph links the subject, measurement, source and applicable rule. Verification and policy checks determine whether the evidence may support the task; AI explains the accepted facts, while the responsible participant controls the operation. A valid signature does not establish truth. Missing evidence can require HOLD. This is an illustrative workflow, not a validated safety case or a reported industrial deployment.
An assertion passes interpretation, authority, status and permitted-use checks before supporting an assurance argument; action authorization remains separate.
Evidence admission preserves the original assertion and checks its interpretation, issuer authority, status and permitted use before it supports a bounded argument. Authorized action remains a separate responsibility. The diagram is an analytical design, not a claim that a signature or graph proves a physical system safe. Status and profile changes may require reassessment of dependent decisions.
Comparison of operational control and dependency dimensions in platform integration and proposed cooperation across independent implementations.
The strategic comparison is platform-governed integration versus participant-controlled cooperation across implementations. Both require meaning, evidence, policy, actions and lifecycle management, and both retain infrastructure and supply-chain dependencies. P1 is illustrated through documented Palantir capabilities, not a claim about all US Big Tech products. P2 is the proposed global cooperation model, building on European dataspaces and international contributions. Its differentiator is a requirement to demonstrate preserved behavior and credible substitution, not an assumed economic or security advantage.
Proposed European, US and Asia-Pacific cooperation through shared exchange contracts while regional participants retain authority and jurisdictional responsibilities.
Proposed global cooperation through open exchange contracts and participant-controlled gateways. European dataspace experience, US innovation and Asia-Pacific initiatives can contribute to shared meaning, discovery and trust under joint governance. Regional participants retain local authority and applicable jurisdictional requirements. Technical interoperability, accepted issuer authority and legal recognition require separate evidence; arrows represent permitted exchanges and governance relationships, not unrestricted data movement or a completed international deployment.
Europe and international partners contribute standards and interoperability profiles while participants retain authority; connecting lines represent a proposed cooperation model.
Europe can turn its dataspace experience into a globally collaborative offer. International partners contribute to shared standards and jointly governed interoperability profiles while participants retain local authority. The illustration is a proposed cooperation model. Diplomatic agreements, technical interoperability, accepted issuer authority and legal recognition are separate achievements; their completion is not implied by the connecting lines.
Both architecture patterns follow corresponding discovery, authorization, verification, decision, action and update steps in a proposed synthetic experiment.
The same discovery, request, authorization, verification, interpretation, decision, action and update responsibilities are assessed in both patterns. The fixture and its oracle are synthetic. This diagram and the test-case table define a proposed experiment; they do not report comparative execution results.

Build a bounded first-mover experiment

Start with a decision that crosses real organizational boundaries: accepting a component, reviewing a changed test report or assembling a release dossier. Agree the relevant identifiers, vocabulary, evidence requirements and responsibilities. Keep each participant’s source processing and action authority local. Exchange only what the task permits. Then test the complete workflow, including missing permission, stale evidence, conflicting claims and service failure.

Measure correctness and unsupported claims, end-to-end response time, tokens and total cost, evidence freshness, recovery and provider substitution. Include the cost imposed on suppliers and laboratories. Demonstrate that a replaceable provider can change without silently losing meaning, authority, evidence history or required behavior. These tests should guide fast-track dataspace modernisation and procurement.

Build the first interoperable evidence workflow with us. We invite manufacturers, suppliers, laboratories, platform providers, standards communities and public-sector partners to agree one useful decision, a shared vocabulary and an open test protocol. Bring a real use case and help demonstrate cooperation across providers. Collaborative leadership matters because no single company can establish an ecosystem’s meaning, trust and operating rules alone.

Discuss a first-mover collaboration →

What is established—and what remains to test?

Editorial review does not mean external journal peer review or acceptance. The v1.0 releases are version-only releases from academic 0.6 and executive 0.5; no new experiments were added. The graph remains the separate fictional 0.1.0 release.

Sources, methods and research boundaries

The papers combine targeted primary-source review, platform economics, comparative architectural synthesis and a proposed matched industrial experiment. Research cutoff: 1 November 2026. This is not a systematic review of every enterprise AI platform. The unchanged PDFs contain the complete, separate bibliographies (101 academic and 30 executive references).

Creative Commons Attribution 4.0 International

Open research

License and citation

This collection of both version 1.0 papers, original figures, research-page content and fictional teaching data is licensed by its named author under the Creative Commons Attribution 4.0 International License (CC BY 4.0), except for the third-party material identified below. Reuse must credit every named author, link to this canonical version and the license, and indicate whether changes were made.

Third-party rights: Referenced third-party material retains its own terms. © 2026 Dr. Carsten Stöcker, Spherity GmbH.

A European Answer to Palantir

Stöcker, C. (2026). A European Answer to Palantir: A globally collaborative counter model for industrial AI (Version 1.0). Spherity GmbH. https://spherity.github.io/spherity-research/European_Industrial_AI_Executive_Brief_v1.0.pdf

A Global Cooperation Model for Sovereign Industrial AI

Stöcker, C. (2026). A Global Cooperation Model for Sovereign Industrial AI: European dataspaces and open standards for verifiable evidence across platforms (Version 1.0). Spherity GmbH. https://spherity.github.io/spherity-research/Industrial_AI_Academic_Paper_v1.0.pdf

Direct answers

Questions this research answers

What is Europe’s answer to Palantir?

This research proposes industrial cooperation across independently governed platforms, using shared semantics, access control and verifiable evidence. Europe’s dataspace experience can help convene that effort globally. It is a design and modernization agenda, not the announcement of a completed European substitute.

Does Palantir only work inside one company?

No. Palantir documents multi-organization ecosystems. The comparison concerns platform-centred integration versus cooperation across independently governed implementations. It does not assert a universal closed-world semantics or a single-company limit.

Why does the ontology matter for AI?

An ontology makes relevant concepts and relationships explicit. A graph can connect a specific component, configuration, report and decision rule so software and people can inspect the evidence. The result can improve grounding and repeatability, provided the mappings, sources and controls are maintained.

Do knowledge graphs eliminate hallucinations or always make AI faster?

No. They can reduce irrelevant context and help check support for claims. Incorrect links, weak sources or unsupported model explanations remain possible. Measure the complete workflow, including graph construction, retrieval and verification costs.

Does open-world semantics mean open access to company data?

No. Open-world semantics treats knowledge as potentially incomplete. Access control decides who may discover or use particular records, relationships and results. A shared vocabulary does not grant permission to read the data that use it.

Are today’s dataspaces already ready for AI-first infrastructure?

That is not established by this research. Current components need workload-specific tests for throughput, latency, volume, evidence freshness and authorization. The papers argue for fast-track modernisation that meets performance and trust requirements together.

What does the robot graph demonstrate?

It shows how fictional evidence can support a configured product and how a withdrawn report can trigger reassessment of dependent decisions. It does not verify signatures, certify a robot or decide whether a regulation applies.

Can one graph automate compliance with every regulation?

Selected evidence can be reused across assessments, but each has its own applicability, authority, evidence and decision criteria. The demonstration separates seven assessment lenses and does not produce a universal compliance approval.

How can companies participate?

Bring a bounded cross-company decision, the relevant participants and a willingness to agree vocabulary, evidence rules and open tests. Spherity invites first movers to test useful cooperation and provider substitution against measurable outcomes.